Wearable devices for protecting against musculoskeletal injuries and enhancing performance

Abstract

To provide a wearable device that protects a person from musculoskeletal injuries and improves performance; specifically, to provide a wearable device that assists a human in performing a physical activity, such as performing an action (e.g., lifting) or maintaining a static posture (e.g., leaning or holding a tool while working on the head).SOLUTION: Materials, constructions, and system architectures allow the wearable devices to be worn over, under, or integrated into clothing for extended periods of time to improve performance or reduce risk of injury. Sensors may be included in the wearable devices to detect various activities, motions, and postures of the wearer, and various active and semi-active controls approaches may leverage sensor information to provide tailored assistance to individual users. Various controls optimization techniques ensure the wearable devices operate at the highest efficiency.SELECTED DRAWING: Figure 8

Description

[Technical field]

[0001] This application claims the benefit of and priority to U.S. Provisional Patent Application No. 62 / 631,666, filed February 17, 2018, and U.S. Provisional Patent Application No. 62 / 757,138, filed November 7, 2018, all of which are incorporated by reference in their entireties herein.

[0002] As part of daily activities, people perform physically demanding tasks that can result in fatigue and increase the risk of musculoskeletal injuries. For example, workers in the manufacturing, construction, warehouse, and logistics industries routinely perform physically demanding activities such as repetitive lifting, lifting heavy loads, transporting equipment and materials over long distances, and maintaining various postures for extended periods of time. These activities can lead to recurrent and disabling injuries that reduce work function and quality of life. In the United States, excessive heavy work accounts for the majority of all nonfatal workplace injuries, costing U.S. businesses more than $13 billion annually in direct costs. Thus, there is an urgent economic and public health need for solutions that can keep our workforce safe and productive.

[0003] As another example, reduced physical strength in older adults can limit and reduce their ability to carry out activities of daily living: sit-to-stand movements, shopping, gardening, etc., are often difficult to perform and lead to fatigue and musculoskeletal injuries.

[0004] A person may need to perform tasks that require more strength or greater effort than they are used to or trained for, which may result in injury, pain, or discomfort (e.g., installing furniture or appliances in a home, moving to a new location, shopping, and carrying heavy equipment).

[0005] Thus, wearable robotic systems that reduce physical strain, mitigate risk of injury, and / or improve the wearer's performance would have a wide range of impacts, including industrial, consumer, geriatric, and medical applications. Soft wearable robots may be used as part of personal protective equipment or integrated into everyday clothing, not only to mitigate the risk of injury during physically demanding activities, but also to be completely transparent so as not to restrict or limit the performance of one person while performing another activity. Summary of the Invention [Problem to be solved by the invention]

[0006] Wearable devices for protecting humans from musculoskeletal injuries and improving performance are disclosed. Systems and methods provide wearable devices that assist humans in performing physical activities, such as performing movements (e.g., lifting) or maintaining stationary positions (e.g., crouching or holding tools while working overhead). [Means for solving the problem]

[0007] The present disclosure relates to a wearable device including a first anchor member configured for placement on a first body part of a person wearing the wearable device; a second anchor member configured for placement on a second body part of the person wearing the wearable device; and at least one actuator directly or indirectly coupling the first anchor member to the second anchor member, where actuation of the at least one actuator generates a tension in the wearable device to generate a moment about one or more joints of the wearer. In one embodiment, the first body part includes an upper body of the wearer and the second body part includes a lower body of the wearer; the one or more joints of the wearer include at least one of (i) one or more hip joints and (ii) one or more back joints; and where the first body part includes at least one shoulder of the wearer and the second body part includes at least one thigh of the wearer, where the one or more joints of the wearer include at least one of (i) one or more hip joints and (ii) one or more back joints. In one embodiment, the first body part includes the wearer's torso and the second body part includes the wearer's upper arm, where the one or more joints of the wearer include at least one shoulder joint. In some embodiments, the first body part includes the wearer's upper leg and the second body part includes the wearer's lower leg, where the one or more joints of the wearer include at least a knee joint. In one embodiment, the first body part includes the wearer's hips and the second body part includes the wearer's upper body, where the one or more joints of the wearer include at least one or more back joints, and further includes a semi-rigid component extending between the first anchor member and the second anchor member, the semi-rigid component being flexible in bending but resistant to deformation under a compressive force.

[0008] In some embodiments, the first body part includes an upper body of the wearer, the second body part includes a first foot of the wearer, and further includes a third anchor member configured for placement on the second foot of the wearer, where actuation of the at least one actuator generates tension in the wearable device and generates a moment about at least one of (i) the hip joint of the first foot and the hip joint of the second foot of the wearer, and (ii) one or more dorsal joints of the wearer; and the embodiments further include a load balancing connector coupling the actuator to the second anchor member and the third anchor member, the load balancing connector including a flexible elongated member having a first end coupled to the second anchor member and a second end connected to the third anchor member, and a mechanism coupling the actuator to an intermediate portion of the flexible elongated member between the first end and the second end, configured to allow the flexible elongated member to translate within the mechanism, thereby balancing the tension distributed to the first foot and the second foot of the wearer. In one embodiment, where the first body part includes an upper body of the wearer, the second body part includes a first foot of the wearer; and further includes a third anchor member configured for placement on a second foot of the wearer, where actuation of the at least one actuator generates tension in the wearable device and generates a moment about at least one of: (i) a hip joint of the first foot and a hip joint of the second foot of the wearer, and (ii) one or more spinal joints of the wearer; this embodiment and further includes a decoupling mechanism integrated into the actuator, The separation mechanism includes a gear configured to engage an output of the actuator motor and rotate when the motor is actuated; a first rotating member coaxially positioned with a first side of the gear and extending from the first side; and a second rotating member coaxially positioned with a second side of the gear and extending from the second side; the first rotating member and the second rotating member are configured to rotate independently of one another and to distribute torque generated by the motor evenly between the second anchor member and the third anchor member, thereby balancing the tension distributed to the wearer's first foot and second foot.In one embodiment, where the first body part includes an upper body of the wearer and the second body part includes a first foot of the wearer; actuation of at least one actuator directly or indirectly coupling the first anchor member and the second anchor member generates a tension in the wearable device to generate a moment about at least one of (i) the hip joint of the wearer's first foot and (ii) the wearer's spinal joint; further comprising a third anchor member configured for placement on the wearer's second foot and at least one actuator directly or indirectly coupling the first anchor member to the third anchor member; actuation of the at least one actuator directly or indirectly coupling the first anchor member and the third anchor member generates a tension in the wearable device to generate a moment about at least one of (i) the hip joint of the wearer's second foot and (ii) the wearer's spinal joint. In one embodiment, the first body part includes a hip of the wearer, the second body part includes a first foot of the wearer, and further includes a third anchor member configured for placement on the second foot of the wearer, at least one actuator directly or indirectly couples the first anchor member to the third anchor member, and actuation of the at least one actuator generates tension in the wearable device and generates a moment about the hip joint of the wearer's first foot and about the hip joint of the wearer's second foot. In one embodiment, the first body part includes a hip of the wearer, the second body part includes a first foot of the wearer, and further includes a third anchor member configured for placement on the second foot of the wearer, at least one actuator directly or indirectly couples the first anchor member to the third anchor member, and actuation of the at least one actuator generates tension in the wearable device and generates a moment about the hip joint of the wearer's first foot and about the hip joint of the wearer's second foot.

[0009] In some embodiments, an actuator is releasably coupled to the first anchor member and the second anchor member, the actuator being selectively attachable to and detachable from the wearable device. In one embodiment, the device further includes a mounting plate coupled to at least one of the first anchor member and the second anchor member, the mounting plate including one or more electrical connectors that place one or more electrical components of the actuator in electrical communication with one or more electrical components associated with at least one of the first anchor member and the second anchor member.

[0010] In one embodiment, the at least one actuator includes a first planar member coupled to a first anchor member, a second planar member coupled to a second anchor member, a plurality of guides disposed on the first and second planar members, and a flexible elongated element that passes alternately through opposing guides on the first and second planar members to couple the first and second planar members, a first end of the flexible elongated element fixed to the first planar member and a second end of the flexible elongated element coupled to a motor, wherein actuation of the motor pulls the second end of the flexible elongated element, thereby pulling the first and second anchor members toward each other.

[0011] In one embodiment, the at least one actuator includes a plurality of guides disposed on the first anchor member and the second anchor member, and a flexible elongate element that alternately passes through opposing guides on the first anchor member and the second anchor member to couple the first anchor member and the second anchor member, a first end of the flexible elongate element fixed to the first anchor member and a second end of the flexible elongate element coupled to a motor, and actuating the motor pulls the second end of the flexible elongate element, thereby pulling the first anchor member and the second anchor member towards each other.

[0012] In one embodiment, at least one actuator includes a motor, a first rotating pulley element connected to the output of the motor, a second rotating pulley element located in the same plane as the first rotating pulley element and radially offset and coupled to the first rotating pulley element via a timing belt, a third rotating pulley element located coaxially and axially offset from the first rotating pulley and fixedly coupled to the first rotating pulley, and a flexible elongated element wrapped around the third rotating pulley, an end or intermediate portion of the flexible elongated element is coupled to the second anchor member, and upon actuation of the motor, the flexible elongated member shortens, thereby pulling the first anchor member and the second anchor member towards each other.

[0013] In another aspect, the present disclosure relates to a wearable device including a first anchor member configured for placement on a first body part of a person wearing the wearable device, a second anchor member configured for placement on a second body part of the person wearing the wearable device, at least one connection element directly or indirectly coupling the first anchor member to the second anchor member, and at least one mechanism configured to selectively fix a length, vary a length, or vary a load on the at least one connection element to control a level of tension generated in the wearable device by a movement or posture of the wearer such that the wearable device generates a moment about one or more joints of the wearer. In one embodiment, the first body part includes an upper body of the wearer, the second body part includes a lower body of the wearer, and the one or more joints of the wearer include at least one of (i) one or more hip joints, and (ii) one or more back joints. In one embodiment, the first body part includes a torso of the wearer, the second body part includes an upper arm of the wearer, and the one or more joints of the wearer include at least a shoulder joint. In some embodiments, the first body part includes an upper leg of the wearer, the second body part includes a lower leg of the wearer, and the one or more joints of the wearer include at least a knee joint. In one embodiment, the first body part includes a hip of the wearer, the second body part includes an upper body of the wearer, and the one or more joints of the wearer include at least one or more back joints, and further includes a semi-rigid component extending between the first anchor member and the second anchor member, the semi-rigid component being flexible in bending but resistant to deformation under a compressive force.

[0014] In another aspect, the present disclosure relates to a wearable device including a first anchor member configured for placement on a first body part of a person wearing the wearable device, a second anchor member configured for placement on a second body part of the person wearing the wearable device, at least one connection element configured to directly or indirectly couple the first anchor member to the second anchor member and absorb energy generated by a movement or posture of the wearer; and at least one mechanism configured to selectively lock the at least one connection element to store the absorbed energy and unlock the at least one connection element to release the absorbed energy such that the wearable device generates a moment about one or more joints of the wearer. In one embodiment, the first body part includes an upper body of the wearer, the second body part includes a lower body of the wearer, and the one or more joints of the wearer include at least one of: (i) one or more hip joints, and (ii) one or more back joints. In one embodiment, the first body part includes the wearer's torso, the second body part includes the wearer's upper arms, and the one or more joints of the wearer include at least one shoulder joint. In one embodiment, the first body part includes the wearer's upper legs, the second body part includes the wearer's lower legs, and the one or more joints of the wearer include at least one knee joint. In one embodiment, the first body part includes the wearer's hips, the second body part includes the wearer's upper body, and the one or more joints of the wearer include at least one back joint, and further includes a semi-rigid component extending between the first anchor member and the second anchor member, the semi-rigid component being flexible in bending but resistant to deformation under compressive forces.

[0015] In another aspect, the present disclosure relates to a wearable device including a first anchor member configured for placement on a first body part of a person wearing the wearable device, a second anchor member configured for placement on a second body part of the person wearing the wearable device, and at least one passive element directly or indirectly coupling the first anchor member to the second anchor member such that a movement or posture of the wearer generates a tension in the wearable device and the wearable device provides a moment about one or more joints of the wearer. In one embodiment, the first body part includes an upper body of the wearer and the second body part includes a lower body of the wearer, and the one or more joints of the wearer include at least one of (i) one or more hip joints and (ii) one or more back joints.

[0016] In some embodiments, the first body part includes the wearer's torso, the second body part includes the wearer's upper arms, and the one or more joints of the wearer include at least a shoulder joint. In one embodiment, the first body part includes the wearer's upper legs, the second body part includes the wearer's lower legs, and the one or more joints of the wearer include at least a knee joint. In one embodiment, the first body part includes the wearer's hips, the second body part includes the wearer's upper body, and the one or more joints of the wearer include at least one or more back joints, and further includes a semi-rigid component extending between the first anchor member and the second anchor member, the semi-rigid component being flexible in bending but resistant to deformation under compressive forces.

[0017] In another aspect, the present disclosure relates to a wearable device including at least one actuator configured to generate forces within or cause the wearable device to generate forces such that the wearer generates moments about one or more joints of the wearer to assist the wearer in performing a non-periodic motion or maintaining a posture; at least one sensor configured to measure information for evaluating an objective function related to providing at least one physical assistance to the wearer, an interaction between the wearer and the wearable device, and an operation of the wearable device; and at least one controller configured to actuate the at least one actuator according to at least one actuation profile, evaluate the objective function based on the information measured by the at least one sensor to determine a resultant change in the objective function, adjust at least one parameter of the at least one actuation profile based on the resultant change in the objective function, and continue the actuation, evaluation, and adjustment to optimize the at least one actuation parameter to maximize or minimize the objective function.

[0018] In one embodiment, the at least one controller is configured to identify two or more operating parameters to be optimized to maximize or minimize an objective function, operate the at least one actuator according to two or more operating profiles having different sets of baseline values ​​for the two or more operating parameters to be optimized, evaluate the objective functions for each of the two or more operating profiles based on information measured by the at least one sensor, define a mathematical correlation between the two or more operating parameters and the corresponding evaluations of the objective functions based on the corresponding evaluations of the objective functions, evaluate the mathematical correlation of the baseline values ​​to determine a candidate set of two or more operating parameter values ​​for maximizing or minimizing the objective function, update the mathematical correlation based on the corresponding evaluations of the objective function for operation of the at least one actuator according to the operating profile associated with the candidate set of the two or more operating parameter values, and continue updating the mathematical correlation until the evaluation of the objective function reaches a global maximum or minimum or a termination criterion is met.

[0019] In another aspect, the present disclosure relates to a wearable device including a first anchor member configured for placement on a first body part of a person wearing the wearable device; a second anchor member configured for placement on a second body part of the person wearing the wearable device; at least one connection element directly or indirectly coupling the first anchor member to the second anchor member; at least one sensor configured to measure information regarding one or more of an angle, a velocity, and an acceleration of one or more joints of the wearer spanned by the wearable device; and at least one controller configured to detect an initiation and / or a type of movement or a posture of the wearer assisted by the wearable device, determine a desired tension to be generated by the wearable device as a function of a given angle, velocity, or acceleration of the one or more joints spanned by the wearable device to assist the wearer in performing the movement or maintaining the posture, and adjust a load on the wearable device or a force provided by the wearable device such that a desired tension is generated in the wearable device at the given angle, velocity, or acceleration of the one or more joints.

[0020] In some embodiments, the wearable device includes at least one mechanism configured to selectively fix a length, change a length, or change a load of at least one connection element, where adjusting the load of the wearable device includes fixing a length of the at least one connection element at a length configured to generate a desired tension in the wearable device with a wearer's movement or posture at a given angle, velocity, or acceleration of one or more joints. In one embodiment, the wearable device includes at least two connection elements, and the wearable device includes at least one mechanism configured to selectively engage one or more of the at least two connection elements, where adjusting the load of the wearable device includes engaging one or more of the at least two connection elements having a spring constant or damping constant, alone or in combination, and configured to generate a desired tension in the wearable device with a wearer's movement or posture at a given angle, velocity, or acceleration of one or more joints.

[0021] In one embodiment, the wearable device includes at least one actuator, and adjusting the load of the wearable device includes actively actuating at least one connection element to generate a desired tension in the wearable device at a given angle, velocity, or acceleration of one or more joints. In one embodiment, the controller is configured to monitor information regarding one or more of the angle, velocity, and acceleration of one or more joints to detect when one or a combination of the angle, velocity, or acceleration exceeds a threshold indicative of a beginning of an assisted movement or posture. In one embodiment, the controller is configured to monitor information regarding one or more of the angle, velocity, and acceleration of one or a combination of the angle, velocity, or acceleration exceeds a threshold indicative of an end of an assisted movement or posture.

[0022] In one embodiment, the assisted movement or posture is a lifting movement or a squatting posture, and the controller determines a relative angle of one or more of the wearer's torso, thigh joint, and hip joint, and the controller monitors the relative angle to detect when the relative angle exceeds a threshold indicative of the start of an assisted lifting movement or squatting posture. In one embodiment, the assisted movement or posture is a lifting movement or a squatting posture, and the controller determines an average angle of the wearer's hip joint, and the controller monitors the average angle of the wearer's hip joint to detect when the average angle of the hip joint exceeds a threshold indicative of the start of an assisted lifting movement or squatting posture. In one embodiment, the controller determines the average angle of the wearer's hip joint by calculating the relative angle between (i) the wearer's torso and (ii) the wearer's thigh.

[0023] In one embodiment, the controller is configured to identify whether the wearer is engaged in an activity other than an assisted lifting or squatting motion, and if so, determine that any excess over a threshold does not indicate the start of a lifting or squatting motion. In one embodiment, the controller determines whether the wearer is engaged in an activity other than an assisted lifting or squatting motion by determining a difference between an angle of one hip joint and an angle of the other hip joint, applies a penalty term to the difference, subtracts the penalty difference from an average hip joint angle, and evaluates whether the resulting determination of the average hip joint angle exceeds a threshold. In one embodiment, the assisted motion or posture is a lifting or squatting posture, and the controller determines an average hip joint angle of the wearer, and further, the controller monitors the average hip joint angle of the wearer to detect when the average hip joint angle exceeds a threshold indicating the end of the assisted lifting or squatting posture. In one embodiment, the controller is configured to monitor information regarding one or more of the angles, velocities, and accelerations of one or more joints to detect one or more states of the assisted motion.

[0024] In one embodiment, the assisted motion is a lifting motion and the controller determines an average hip angle of the wearer, and the controller monitors the average hip angle of the wearer to detect when the average hip angle of the wearer exceeds a threshold indicative of a beginning or ending phase of the lifting motion. In one embodiment, the threshold is indicative of an initial down state of the lifting motion and the threshold is indicative of an average hip angle increasing from a neutral hip angle. In one embodiment, the threshold is indicative of a hold state of the lifting motion and the threshold is an average hip angular velocity decreasing to about 0 degrees per second. In one embodiment, during the hold state, the controller monitors the average hip angle to detect a transition to a lift state or a down state, where a change in the average hip angle to a smaller angle indicates a transition to the lift state and a change in the average hip angle to a larger angle indicates a transition to the down state.

[0025] In one embodiment, the assisted motion is a lifting motion, and the controller is configured to classify the lifting motion as a bend lifting motion, a squat lifting motion, or a bending or twisting lifting motion based on information related to one or more of the angles, velocities, and accelerations of one or more joints. In some embodiments, the controller is configured to adjust the load of the wearable device as a function of the state or type of the assisted motion or posture. And during a holding state, the controller is configured to progressively increase the tension to a target value to assist the wearer in holding the posture for an extended period of time. And the controller is configured to receive input from the wearer to select a peak tension to assist the different motions, and further configured to scale the load or force provided by the wearable device to have a maximum peak tension corresponding to the selected peak tension value. And the controller is configured to receive input from the wearer to disable generation of tension by the wearable device.

[0026] In another aspect, the present disclosure relates to a wearable system for monitoring a person's physical activity, the system including at least one sensor configured to measure information regarding one or more of the following: an angle, velocity, or acceleration of one or more joints spanned by the wearable system of the wearer, a force or torque experienced by the wearer, and a biometric parameter of the wearer; and at least one processor configured to analyze the information measured by the at least one sensor to assess one or more of injury risk, work capacity, biomechanics, and ergonomics metrics associated with the wearer's physical activity while wearing the wearable system. In one embodiment, the at least one sensor includes one or a combination of an inertial measurement unit (IMU), a joint angle sensor, a force or pressure sensor, a torque sensor, a metabolic energy measuring device, an muscular activity measuring device (EMG), a ground reaction force sensor, a heart rate sensor, and an insole force or pressure sensor. In one embodiment, the processor is further configured to notify the wearer that the wearer is performing a motion that may lead to injury. In one embodiment, the processor is further configured to determine and advise the wearer of an appropriate recovery period to mitigate potential injury before engaging in further activity. [Brief description of the drawings]

[0027] [Figure 1] It presents a variety of application areas where users can benefit from wearing assistive devices to reduce the risk of musculoskeletal injuries and increase productivity.

[0028] [Figure 2A] 1 shows a sample embodiment of a passive device. [Figure 2B] 1 shows a sample embodiment of an active device for supporting the body including the back and hip joints. [Figure 2C] 1 shows a sample embodiment of a sensor-only system.

[0029] [Diagram 3] 1A-1C show multiple views of an exemplary embodiment of a robotic garment.

[0030] [Figure 4A] 13 shows how the output tension can be modified by defining different initial lengths of the connecting elements. [Figure 4B] 13 shows how the output tension can be modified by defining different initial lengths of the connecting elements.

[0031] [Figure 5A] Shown is a sample user motion in which the joints follow a sinusoidal pattern, moving from an initial position of 0° (standing) to 90° (bending the torso into a horizontal position). [Figure 5B] Shown is a sample user motion in which the joints follow a sinusoidal pattern, moving from an initial position of 0° (standing) to 90° (bending the torso into a horizontal position). [Figure 5C] Shown is a sample user motion in which the joints follow a sinusoidal pattern, moving from an initial position of 0° (standing) to 90° (bending the torso into a horizontal position).

[0032] [Figure 6] It is shown how two different healthy individuals can have different outcomes depending on the chosen actuation parameters (in this case actuation timing), with the maximum capacity of the system being achieved by very different actuation parameters for each subject.

[0033] [Figure 7] We show how actuation profiles can be decomposed according to given timing, speed, magnitude, and overall actuation shape.

[0034] [Figure 8] We present representative methods for optimal control of wearable devices.

[0035] [Figure 9]We show an IMU being used to detect the initiation of movement (changes in joint acceleration or velocity) and the direction of movement, e.g., bending and / or twisting of the torso, in order to define and classify the movement as whether the user is performing the movement in the sagittal plane (e.g., lifting an object in front of the user) or whether the torso is bending and twisting (e.g., lifting an object sideways).

[0036] [Figure 10] 13A-13C show an embodiment of a robotic suit in which anchor members span multiple joints.

[0037] [Figure 11] The proposed device shows several industries with multiple application areas. [Figure 12] The proposed device shows several industries with multiple application areas.

[0038] [Figure 13] An example of a load balancing strap connecting both thigh anchor members to create an anchor point at one of the dorsal points.

[0039] [Figure 14] Demonstrating the functionality of the load balancing straps.

[0040] [Figure 15] It is shown how the output tension can be altered by defining different initial lengths of the connecting elements, and how different stiffness values ​​of the connecting elements provide different outputs. [Figure 16] It is shown how the output tension can be altered by defining different initial lengths of the connecting elements, and how different stiffness values ​​of the connecting elements provide different outputs.

[0041] [Figure 17]1 shows a device as described in Example #1B providing supportive tension to the back and hip joints that is passively generated by the user's body movement. The connection elements may be sized or designed to provide different types of support as described in Example 1A.

[0042] [Figure 18] A device is presented that includes an actuator component that can be used to control a cable attached between both anchor members.

[0043] [Figure 19] We show how a sample hypothetical load provided by actuation can be defined in terms of stiffness values: where F is the resulting force, θ is the relative angle between the thigh and torso, and K is the provided stiffness.

[0044] [Figure 20A] Shown is a sample user motion where the joints are moving according to a sinusoidal pattern from an initial position of 0° (standing) to 90° (bending the torso into a horizontal position). [Figure 20B] Shown is a sample user motion where the joints are moving according to a sinusoidal pattern from an initial position of 0° (standing) to 90° (bending the torso into a horizontal position). [Figure 20C] Shown is a sample user motion where the joints are moving according to a sinusoidal pattern from an initial position of 0° (standing) to 90° (bending the torso into a horizontal position).

[0045] [Figure 21] An example of a semi-rigid part is shown that includes two anchor members, one at the top and one at the bottom. By connecting connecting elements such as actuation cables or elastic passive elements, an assist torque can be provided to the user's back.

[0046] [Figure 22]An example of a soft wearable device embodiment is shown, which is achieved by placing a cable-driven actuator with anchor members on either side of the knee (front of the thigh and front of the calf).

[0047] [Diagram 23] A configuration is shown in which the actuator has anchors under the arms attached to garment pieces that wrap around the arms and around the sides of the chest so that as the cables move they generate tension that creates a supporting torque at the shoulder joint.

[0048] [Figure 24A] We show how the devices described in the previous examples can be combined to provide support for multiple joints, such as the knee, back, shoulder, or any combination thereof. [Figure 24B] We show how the devices described in the previous examples can be combined to provide support for multiple joints, such as the knee, back, shoulder, or any combination thereof. [Figure 24C] We show how the devices described in the previous examples can be combined to provide support for multiple joints, such as the knee, back, shoulder, or any combination thereof.

[0049] [Figure 25A] Some sample embodiments of how the proposed device can be integrated into typical work clothing for everyday use are shown. [Figure 25B] Some sample embodiments of how the proposed device can be integrated into typical work clothing for everyday use are shown. [Figure 25C] Some sample embodiments of how the proposed device can be integrated into typical work clothing for everyday use are shown. [Figure 26A] Some sample embodiments of how the proposed device can be integrated into typical work clothing for everyday use are shown. [Figure 26B] Some sample embodiments of how the proposed device can be integrated into typical work clothing for everyday use are shown.

[0050] [Figure 27] 1 shows a wearable support device that is worn around the shoulder to support the shoulder joint when the user extends their arms over their head, preventing the shoulder joint from overexerting.

[0051] [Figure 28] An example flowchart of an optimization process is shown, which includes evaluating objectives based on initial settings or search points, updating profiles using collected data, determining next candidate search points, evaluating objectives at these settings, checking convergence criteria, etc.

[0052] [Figure 29A] Showcasing a range of activities that the device can optimise and assist with.

[0053] [Figure 29B] A classification algorithm is presented that classifies the different types of activities that the user is performing during the initialization phase of the operation, allowing to select different assistance parameters for each operation and optimize them independently.

[0054] [Diagram 30] Small to moderate levels of assistance (20%-30% of typical joint torques during comfortable walking) combined with the low inertia and non-restrictive nature of the proposed robotic clothing platform provide a further example of existing new opportunities for the control of wearable robots where assistance needs to be synchronized with the wearer's underlying muscle function.

[0055] [Diagram 31] Examples of representative support profiles by varying the parameter k are shown. Higher stiffness values ​​mean more support for the back as a function of trunk angle during bending movements.

[0056] [Diagram 32]FIG. 1 is an example of an assist device that may include one actuation unit on the back of each finger that can provide assistance to each finger when pressed.

[0057] We represent how the robotic garment is load-controlled; we are developing a commercially available passive rigid exoskeleton and a soft passive device to offload the back and hip joints. Our preliminary results show that for some users, passive loading is effective in offloading the targeted back muscles for some movements.

[0058] [Figure 34A] SUMMARY OF THE DISCLOSURE An example transmission system and methodology is presented. [Figure 34B] SUMMARY OF THE DISCLOSURE An example transmission system and methodology is presented. [Figure 34C] SUMMARY OF THE DISCLOSURE An example transmission system and methodology is presented.

[0059] [Diagram 35] We show that a robotic garment can be designed to be worn underneath clothing and consist of (i) a fully soft undergarment containing soft sensing components and force paths for comfortable transmission of assistive forces to the body, and (ii) a cable-driven actuation module that meets specific force, power and bandwidth requirements and can be easily connected to the undergarment to power the distributed sensors and assist the user.

[0060] [Diagram 36] Illustrates the concept of a device that may be worn to assist a user's movement or posture during lifting and / or reaching tasks.

[0061] [Figure 37A] FIG. 1 shows a diagram illustrating how tension can generate assistive torque across the spinal and hip joints.

[0062] [Figure 37B]An embodiment of a fully portable device capable of supporting the back and hip joints is shown according to the structure shown in FIG. 37A.

[0063] [Figure 38] 1 shows a cable driven actuation system that can be connected between two anchor members.

[0064] [Figure 39A] 13 shows a design drawing of an alternative cable driven actuation system. As shown, this actuation element includes a quick attachment mechanism for the wearable device that allows for the design attachment of the textile part to the machining mechanism. [Figure 39B] 13 shows a design drawing of an alternative cable driven actuation system. As shown, this actuation element includes a quick attachment mechanism for the wearable device that allows for the design attachment of the textile part to the machining mechanism. [Figure 40A] 13 shows a design drawing of an alternative cable driven actuation system. As shown, this actuation element includes a quick attachment mechanism for the wearable device that allows for the design attachment of the textile part to the machining mechanism. [Figure 40B] 13 shows a design drawing of an alternative cable driven actuation system. As shown, this actuation element includes a quick attachment mechanism for the wearable device that allows for the design attachment of the textile part to the machining mechanism. [Figure 41A] 13 shows a design drawing of an alternative cable driven actuation system. As shown, this actuation element includes a quick attachment mechanism for the wearable device that allows for the design attachment of the textile part to the machining mechanism. [Figure 41B] 13 shows a design drawing of an alternative cable driven actuation system. As shown, this actuation element includes a quick attachment mechanism for the wearable device that allows for the design attachment of the textile part to the machining mechanism.

[0065] [Diagram 42]A device including a cable-driven electromechanical actuation system is shown being worn by a user.

[0066] [Figure 43A] Shows further details of the actuator system positioned on the user's torso and its integration into the garment. [Figure 43B] Shows further details of the actuator system positioned on the user's torso and its integration into the garment. [Figure 43C] Shows further details of the actuator system positioned on the user's torso and its integration into the garment.

[0067] [Diagram 44] Various users are shown wearing the proposed system over their clothing to perform various tasks, including handing and lifting packages. [Diagram 45] Various users are shown wearing the proposed system over their clothing to perform various tasks, including handing and lifting packages.

[0068] [Figure 46A] In one embodiment, the components, structures, and actuator systems of the wearable device may be incorporated into typical clothing for everyday use. [Figure 46B] In one embodiment, the components, structures, and actuator systems of the wearable device may be incorporated into typical clothing for everyday use.

[0069] [Figure 47A] 13 shows further details of an embodiment of integrating robotic parts into clothing. [Figure 47B] 13 shows further details of an embodiment of integrating robotic parts into clothing. [Figure 47C] 13 shows further details of an embodiment of integrating robotic parts into clothing. [Figure 47D]13 shows further details of an embodiment of integrating robotic parts into clothing.

[0070] [Figure 48] An example is shown where the actuator clicks into a mounting plate that is sewn into the waist fabric piece. The mounting plate can have mechanical and electrical mechanisms for clicking in.

[0071] [Figure 49] Shown are sensors designed to be compatible with fabrics to measure human motion, strap tension, interaction force / pressure and muscle strength.

[0072] [Figure 50] 1 illustrates an example of a control framework.

[0073] [Figure 51A] We show how to use an Inertial Measurement Unit to measure joint angles, velocities, and accelerations. [Figure 51B] We show how to use an Inertial Measurement Unit to measure joint angles, velocities, and accelerations. [Figure 51C] We show how to use an Inertial Measurement Unit to measure joint angles, velocities, and accelerations.

[0074] [Figure 52A] Relative angles during walking, stair climbing, and lifting tasks are shown. [Figure 52B] Relative angles during walking, stair climbing, and lifting tasks are shown.

[0075] [Figure 53A] Indicates the starting threshold for determining the operating state.

[0076] [Figure 53B] An example is given in which an initial threshold in the displacement angle can be used so that the system does not provide assistance during the initial phase when the user bends the torso or legs forward.

[0077] [Figure 54] A control method is shown that includes two main components: a high-level controller and a low-level controller, each of which generates an output signal based on various sensor inputs and control logic.

[0078] [Figure 55] It is shown that a low level control scheme for controlling the force or tension in a system consists of two parts. An admittance controller and a feedforward model are described. The motor controller implements two cascaded PI loops for speed control and current control.

[0079] [Figure 56] The left plot shows that the stability of static trunk bending (η = 0) depends primarily on the level of support Fd. The right plot shows the minimum damping value required to ensure stable tracking in static forward bending postures. The right plot shows the damping scheduling approach by qualitatively outlining the range of reasonable damping values ​​for controlling the outer garment based on the dynamics factor η. The shaded areas in the right plots represent the ideal damping values ​​determined for each task.

[0080] [Figure 57] The garment behavior is shown to include hysteresis in tensioning (loading) the cable and releasing (unloading) it.

[0081] [Figure 58] Ignoring suit deformations, we show the theoretical relationship between torso motion, thigh motion, and the corresponding cable lengths.

[0082] Feedforward to account for the effect of human movement on cable velocity Demonstrates the abilities of word models.

[0083] [Figure 60A] It is shown that support can be defined as a function of the trunk-thigh angle. [Figure 60B] It is shown that support can be defined as a function of the trunk-thigh angle. [Figure 60C] It is shown that support can be defined as a function of the trunk-thigh angle. [Figure 61A] It shows how different states of lifting and different loads can be defined. [Figure 61B] It shows how different states of lifting and different loads can be defined.

[0084] [Figure 62] 1 is a plot showing measurements from a device where five different users performed a series of 50 lifting cycles and the system provided different types of load profiles.

[0085] [Figure 63A] 1 shows measurements of the applied force and displacement angle from the device following the presented assistance method, where the user began in an upright position and performed a series of four lifts at different speeds and with different ranges. [Figure 63B] 1 shows measurements of the applied force and displacement angle from the device following the presented assistance method, where the user began in an upright position and performed a series of four lifts at different speeds and with different ranges.

[0086] [Figure 64] A device is shown that uses sensory data to detect different motions as the user moves downwards, detecting the user's motion and adapting the assistive force to a profile more appropriate for that particular motion.

[0087] [Figure 65] We show that machine learning algorithms such as decision trees (Figure 65A) and hidden Markov networks (Figure 65B) can be used to classify user actions.

[0088] [Figure 66A] The concept of three-dimensional load space is shown. [Figure 66B] The concept of three-dimensional load space is shown.

[0089] [Figure 67A] An exemplary approach for providing static retention assistance is illustrated, where the control method may increase the amount of assistance depending on the time the wearer remains in the "retention" position. [Figure 67B] An exemplary approach for providing static retention assistance is illustrated, where the control method may increase the amount of assistance depending on the time the wearer remains in the "retention" position. [Figure 67C] An exemplary approach for providing static retention assistance is illustrated, where the control method may increase the amount of assistance depending on the time the wearer remains in the "retention" position. [Fig. 67D] An exemplary approach for providing static retention assistance is illustrated, where the control method may increase the amount of assistance depending on the time the wearer remains in the "retention" position.

[0090] [Figure 68A] 13A-13C illustrate embodiments in which the load balance assembly may further include a hardware component that can slide up and down the fuselage to decouple the movement of the load balance straps from the elements (e.g., cable components) to which they are connected.

[0091] [Figure 68B] In another embodiment, one or more separate components may be incorporated into the actuation system itself.

[0092] [Figure 68C] 1 shows a representative isolation component incorporated into an actuator.

[0093] [Figure 69] Demonstrates the concept of load balancing straps being decoupled from cable components using low friction components, rollers, bearings, pulleys, etc.

[0094] [Figure 70] An alternative way of achieving a load balancing mechanism is described, in this case using two independent motors to actuate each strap / cable individually.

[0095] [Figure 71] 1 illustrates an embodiment in which a wearable device is used.

[0096] [Figure 72] We show that a control method can exploit the fact that the sum of the forces applied to both legs (which exert a downward force on the waist belt) is greater than the back force (which exerts an upward force on the waist belt) and that holding the waist belt in place is easier because the shape of the torso helps to hold the waist belt in place, controlling back support and thigh support independently but limiting the magnitude of back support to a force close to or smaller than the sum of the forces applied to both legs.

[0097] [Figure 73] Show how straps passed around the legs (e.g., harnesses) or semi-rigid parts secured to the waist belt on one side and in a fixed position on the user's shoulders at the upper torso on the other side can provide additional support to the waist belt to avoid upward movement. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0098] A wearable device 100 is disclosed for protecting humans from musculoskeletal disorders and improving performance. Systems and methods provide a wearable device 100 that assists humans in physical activities, such as performing a motion (e.g., lifting) or maintaining a static posture (e.g., bending over or holding a tool while working overhead). Materials, structures, and system configurations allow the wearable device 100 to be worn over, under, or incorporated into clothing for extended periods of time to improve performance or reduce the risk of injury. Sensors 230 may be included in the wearable device 100 to detect various activities, motions, and postures of the wearer. Various active and semi-active control techniques can then leverage the sensor information to provide tailored assistance to individual users. Various control optimization techniques ensure that the wearable device 100 operates at peak efficiency.

[0099] Whether enhancing and protecting workers in various industries (e.g., construction workers, manufacturing, logistics, warehouses, baggage handlers, installers, caregivers, physicians, etc.) or as part of activities of daily living (e.g., seniors, healthy individuals, medical patients), the soft wearable device 100 can have a wide range of impacts to reduce musculoskeletal injuries and improve productivity and / or performance. By way of example, FIG. 1 illustrates a variety of application areas where users may benefit from wearing assistive devices to reduce the risk of musculoskeletal injuries and improve productivity.

[0100] Disclosed are methods, systems, and devices that assist a user when performing physical activities or movements that may lead to discomfort, pain, fatigue, injury, etc. More specifically, the disclosure relates to a wearable device 100 (e.g., a soft wearable garment, an exoskeleton, a wearable robotic device, or a robotic garment) that may be worn for extended periods of time and that includes the following elements or a combination of the following elements: (i) A wearable device structure, comprising anchor members 110 configured for placement on various body parts of a person wearing the wearable device 100, and flexible connection elements 150 (such as clothing, semi-rigid parts, webbing, straps, cables, combinations thereof, etc.) that directly or indirectly couple the anchor members 110. In various embodiments, the connection elements 150 can be actively actuated to actively generate tension in the wearable device 100, semi-actively actuated to vary the level of tension generated in the wearable device 100 with the wearer's movements or posture, and / or passively generate tension in the wearable device 100 when the wearer is based on the wearer's movements or posture. As configured, the wearable device 100 delivers forces and / or moments to the body to assist the wearer in performing various movements or maintaining various postures. (ii) Wearable sensors 230 for measuring or estimating human biomechanics, interactions between the system and the wearer, and interactions with the environment or objects. By way of example, these sensors 230 may be integrated into or compatible with clothing components. (iii) Passive components 190 may be integrated such that tension or torque is generated as the user moves, which may be beneficial to the human body. Examples include elastics, springs, and dampers. (iv) Actuation: One or more actuators 120 may be integrated as part of an active system to assist the user or to control the position, force, or pretension of system components. Additionally, in a semi-active system, one or more actuators 120 may be used to set the level of pretension of the system or to set the initial position of an elastic component to regulate when that component begins to passively provide an assistive force with respect to human movement.

[0101] The proposed system may be worn to assist the wearer by providing tension between anchor members 110 located on either side of a joint (e.g., thigh, back, knee, shoulder, elbow) or bridging multiple joints to simultaneously assist multiple joints, body parts, or transfer loads to different parts of the body. The connecting elements 150 between these anchor members 110 may provide tension to assist the user. Connection parts 150

[0102] In various embodiments, one or more of the connection components 150 are controlled by one or more actuators 120 that can control the position of the components or the tension that the connection components 150 generate to provide customized assistance or different levels of assistance to the wearer (active system). In another embodiment, a passive element 190 can be used as the connection component 150 (e.g., elastic components, dampers, etc.) so that as the user moves, the component 190 generates a force that assists the wearer (passive system). In yet another embodiment, the actuator 120 can be used to set a level of pretension of the passive connection element 190 that sets an initial position or tension relative to the anchor member 110 (semi-active system). This can be useful to allow customization of assistance to human movements (e.g., when assistance starts, stops), compensate for relative movements of the wearable device 100 components over time, or set initial pretension to accommodate different body types. Alternatively, some of these adjustments may be made manually via components such as Velcro® straps, ratchet systems (e.g., Boa®, etc.), or buckles. Another embodiment may not include a connection element 150 at all, such that the system does not provide any assistance to the user, but may collect on-board sensor information. FIGS. 2A, 2B, and 2C show sample embodiments of a passive (FIG. 2A), active device 100 (FIG. 2B), and sensor-only system (FIG. 2C) for assisting the body, including the back and hip joints. It should be understood that in various embodiments, the semi-active system may appear similar to the active system of FIG. 2B, except that the actuator 120 or clutch 160 may be controlled to set pretension or engage / release the elastic component 150.

[0103] 150 actuation connection parts: the connecting element 150 between the anchor points can be actuated. One example is a cable drive system where the actuator 120 can be worn by the user, a Bowden cable connecting the actuator 120 unit to one side of the joint, and an inner cable 150 connecting one end to the actuator 120 and the other end to the other side of the joint. In this way, when the inner cable 150 is actuated, tension is created between both anchor points. In another embodiment, the actuator 120 can be worn and connected directly to one side of the joint, actuating a cable 150 that is connected to the actuator 120 at one end and to the other side of the joint. In another embodiment, the actuator 120 can be a fluid (e.g., pneumatic, hydraulic, etc.) and the connecting element 150 can apply tension when the actuator 120 connecting to both attachment points is pressurized or when a vacuum is applied. The connection elements 150 may be non-extensible, such as cables made of steel that can be controlled by the actuator 120 when high actuator 120 bandwidth is important, in another embodiment the connection elements 150 may have series spring elements, or connection elements 150 made of elastic material. In this case bandwidth is traded off in order to have a system that can measure elastic displacement and correlate it with force, thus eliminating the need to include other beneficial properties such as absorbing and reducing component or environmental vibrations using additional sensors 230, elastic elements, etc. (if the material properties of this elastic element are known).

[0104] Motor Gearing Alternatives and Their Integration into Wearable Device 100 Components: In various embodiments, the actuator 120 may require a gear box or transmission such that a series of gear stages reduce the speed and increase the torque at its output to meet demands. Standard gear boxes have several disadvantages, some examples of which include one or more of the following: added size and weight to the design, cost, susceptibility to shock loads, backlash, inefficiency, noise due to tooth meshing, and other factors.

[0105] We present an alternative solution to overcome the problems of gearing in wearable systems. We propose to use a series of rotating elements (e.g. pulleys, low-friction rollers, etc.) that can be integrated into soft parts of the structure (e.g. textile straps, textile parts) that act as an alternative to a gear box, obtaining the advantages of the gear reduction without some of its drawbacks. Figures 34A, 34B and 34C show an example of the proposed transmission method. Figure 34A shows a standard embodiment where the actuating element 150 is attached between two anchor points. The example shown in Figure 34B or 34C shows how a reduction equivalent to a gear box with a (n+1):1 ratio can be achieved by passing the actuating element 150 around a guide 127 (e.g. a rotating or low-friction element). Here, n is the number of guides 127, and this ratio can be changed by reducing or adding rotating elements (n). Assuming that the friction in the guide 127 is negligible, the entire cable 150 has the same tension T throughout its length. This transmission pulls n+1 parallel cable sections on the strap. This means that each section is pulling the strap with a force T. As a result, the total force pulling the strap is F=(n+1)T. Due to conservation of force, if the linear force increases by a factor of n+1, the speed also decreases by a factor of 4. This is the basis of (n+1):1 reduction. In other embodiments, the motor 121 can be attached to the strap components by an existing cable 150 via the proposed structure (shown in Figure 34B) or include other transmissions such as a Bowden cable that attaches to one component on one side and the other component on the other side via guide 127 (Figure 34C). The proposed transmission is ideal for our use case for several reasons. Its design is low profile and takes advantage of the large surface area of ​​the user as opposed to bulky non-low profile gear boxes. The transmission uses inexpensive pulleys or low friction materials instead of expensive gears that require high tolerances. The lack of gears also allows the proposed transmission to have less backlash under tension, potentially reducing efficiency losses and audible noise associated with meshing gear teeth.This concept can be applied to any actuating element that follows a linear motion (electromechanical, fluid, etc.) by connecting this element via a cable following the design above. This concept is also useful for implementations where there is a single pulley, roller, or low friction element (n=1) and a gear reduction ratio of 2:1. This concept can eliminate the need for an additional gear stage in some applications, significantly reducing cost and complexity.

[0106] Passive Connection Element 190 In various embodiments, the connection element 150 can be passive (e.g., a spring, damper, or flexible element (extensible, semi-extensible, or inextensible)) such that tension between both anchor points is passively generated as the person moves or bends. The passive element 190 is designed to passively generate tension to support one or more joints or body parts of the wearer. The configuration (e.g., initial length) of the element 190 can be adjusted to accommodate the Velcro® (登録商標) , buckles, ratchet systems (e.g. Boa (登録商標) ) or other such straps. Additionally, in semi-active systems, actuators 120 can be used to modify these parameters, for example to set pretension levels.

[0107] Sensor-only systemsIn various embodiments, the connection element 150 may not be present, and the system may utilize the on-board sensors 230 to analyze data from one or more workers, for example, during a workday, to assess injury risk, work capacity, biomechanical, and / or ergonomic metrics. This data may be used to optimize various processes within the industry, to notify users when they are performing movements that may lead to injury, and to assess areas of risk. Notification methods may include low-power feedback (e.g., auditory, tactile, visual, etc.) to notify users. These methods may also help reduce injury risk for athletes, consumer applications, or the elderly. In one example, athletes can use this information to know when their risk of injury is increased, and to use this information to establish recovery periods and practice. In another example, athletes or consumers can use information from these metrics to improve their performance. Examples of wearable sensors 230 include one or a combination of inertial measurement units (IMUs), joint angle sensors 230, force or pressure sensors 230, torque sensors 230, metabolic energy measuring devices, muscle activity measuring devices (EMG), ground reaction sensors 230, heart rate sensors 230, and insole force or pressure sensors 230. These sensors 230 may be integrated into garment components. Anchor member 110

[0108] The structure of the wearable device 100 may be designed to include garment components such as straps that wrap around body parts to transfer loads to the body and create anchor points for the connection elements 150. In one embodiment, the system may be designed to actuate different joints or combinations thereof (e.g., back, shoulder, knee, thigh, elbow). The examples provided below provide a detailed description and design of the system to assist these joints.

[0109] Parts of the wearable device 100 that conform to the body and define the structure of the wearable device 100 may include:

[0110] (a) Femoral anchoring element This component of the wearable device 100 is configured to be placed on the wearer's thigh and provides an anchor point on the thigh to create an attachment point on one side of the hip or knee joint. The component incorporates a non-slip material, is firmly fixed to the leg when force is applied, and is made of a stiff fabric to minimize deformation when force is applied to the component. An exemplary embodiment of a thigh anchor element is a thigh wrap.

[0111] (b) Waist anchor member : A strap that wraps around the waist, defining an anchor point in the body to which the actuator 120 or passive element 190 can be attached. An exemplary embodiment of a lumbar anchor element is a waist belt.

[0112] (c) Shoulder anchor member This component of the wearable device 100 is configured to be placed on the wearer's shoulder and provides a shoulder anchor point to create an anchor point on the upper back. An exemplary embodiment of a shoulder anchor element is a shoulder strap.

[0113] (d) Thigh Strap 112 : an elongated element (e.g., strap, ribbon, cable) that connects to both thigh anchor members 110 to create an attachment point on the underside of the spine joint. The next figure shows how the strap 112, which has each end 113, 114 attached to the back of each thigh anchor member 110 to create a loop that constitutes an attachment point at the waist, can be integrated into the system. In some embodiments, the thigh strap 112 can function as a load balancing element 212 of the load balancing assembly 210, as described in more detail below.

[0114] FIG. 3 shows various views of an exemplary embodiment of a robotic suit that includes some of the components described above, as indicated. Sensor 230

[0115] Various sensors 230 can be used to measure various reactions of the body, its interaction with the wearable system 100 or with the environment. In some embodiments, sensors 230 configured to measure human biomechanics can include, but are not limited to, inertial measurement units (IMUs), accelerometers, gyroscopes, encoders, resolvers, and strain sensors 230 to measure joint angles, velocities, and accelerations, heart rate monitors to measure heart rate, pulmonary gas exchange systems to measure metabolism, electromyography, and pressure-sensing insoles and instrumented treadmills to measure ground reaction forces. Sensors 230 that can be used to measure wearer-system interactions, including forces and pressures, include, but are not limited to, load cells, force sensors 230, and pressure sensors 230. Motion sensors 230 can be used to estimate joint angles and dynamics (angular velocity, acceleration) by placing one motion sensor on each part of the body. As an example, using the IMU on the back, When performing activities such as lifting, bending, maintaining static positions, and going from a sitting to standing position, You can know the angle of the torso,The relative angle between the torso and thigh can be used to clarify whether a person is walking (cyclical movement of both legs) or squatting to lift an object (both legs bent), the relative angle between the shoulder IMU and the forearm IMU can be used to detect when a user is performing an overhead task that requires assistance such as a shoulder assist device, and the IMUs on both sides of the knee joint can be used to estimate the knee angle, velocity, or acceleration to know when a person is in a squatting position that may require assistance from a knee assist device. As an example, a load cell or force sensor embedded in the connection element 150 or between the connection element 150 and the anchor point can be used to know the force the connection element is transmitting to the user. As another example, a pressure sensor placed between the user and the strap can be used to estimate or measure force or pressure for comfort or pressure distribution purposes. As yet another example, sensors 230 embedded in the robotic garment and sensors 230 placed on the user can be used to know the relative motion between the robotic garment and the user. Sensors 230 that can be used to measure metrics of the device include, but are not limited to, current sensors 230, voltage sensors 230, thermistors, and encoders and resolvers to measure motor position, speed, and acceleration. As an example, voltage and current sensors 230 can be used to estimate the power that the actuator 120 is consuming, and as another example, voltage can be used to estimate the battery status. Thermistors can be used to evaluate the thermal state of the system, which can be useful as a prognostic diagnostic, or to define the amount of assistance to stay below a marginal threshold in any type of environment, taking heat into account. Sensors 230 such as motor position, speed, and acceleration are necessary to control the device and can also be effective sensors 230 to evaluate prognostics and condition monitoring. In one embodiment, these sensors 230 may be textile-based and / or compatible with textile elements. As an example, the sensors 230 may be constructed by fabrics that change electrical properties (e.g., resistance, capacitance) when stretched or when pressure is applied.This change can be measured by an electrical circuit and correlated with joint movements, interaction forces between the system and the user, or between the environment. Passive Wearable Devices

[0116] The wearable device 100 of the present disclosure, in various embodiments, may be configured to passively generate a moment about one or more joints of a wearer to assist the wearer in performing a motion or maintaining a posture. A passive wearable device 100 may generally include at least one anchor member 110 configured for placement on a first body part of a person wearing the wearable device 100, at least one anchor member 110 configured for placement on a second body part of the person wearing the wearable device, and at least one connection element 150 that directly or indirectly couples the anchor member 110 of the at least one first body part to the anchor member 110 of the at least one second body part such that a motion or posture of the wearer generates a tension in the wearable device 100 and the wearable device 100 provides a moment (e.g., a restoring torque) about one or more joints spanned by the wearable device 100.

[0117] For example, the exemplary passive wearable device 100 shown in FIG. 2A includes two upper body anchor members 110 (e.g., shoulder anchor members 110 in the form of shoulder straps), two lower body anchor members 110 (e.g., thigh anchor members 110 in the form of thigh wraps), and a passive connection element 190 (e.g., a spring, damper, or flexible element (tensile, semi-tensile, or intensile)) connecting the upper body anchor members 110 to the lower body anchor members 110. As configured, the wearer's movement or posture can generate tension within the wearable device 100 and provide a moment (e.g., restoring torque) about at least one of: (i) one or more hip joints of the wearer, and (ii) one or more back joints of the wearer. This will be described in more detail throughout this disclosure.

[0118] References to back joints in this disclosure are intended to broadly include any and all parts of the torso that may flex or move relative to other parts of the torso, including but not limited to joints such as lumbar joints, sacroiliac joints, and other joints, as well as connections between individual vertebrae. Semi-active wearable devices

[0119] The wearable device 100 of the present disclosure, in various embodiments, may be configured to semi-actively generate moments about one or more joints of the wearer to assist the wearer in performing a motion or maintaining a posture. Generally speaking, a semi-active wearable device 100 does not actively generate moments about one or more joints of the wearer, but instead adjusts the load (e.g., stiffness, damping) of the wearable device 100 such that the wearer's motion or posture generates a desired level of tension within the wearable device 100 to assist the wearer in performing a motion or maintaining a posture. A semi-active wearable device 100 may generally include at least one anchor member 110 configured for placement on a first body part of a person wearing the wearable device 100, at least one anchor member 110 configured for placement on a second body part of the person wearing the wearable device, a connection element 150 that directly or indirectly couples the anchor member 110 of the at least one first body part to the anchor member 110 of the at least one second body part, and at least one semi-active actuator 120 configured to control a load on the wearable device 100 such that a movement or posture of the wearer generates a tension in the wearable device 100. This tension causes the wearable device 100 to generate a moment about one or more joints that the wearable device 100 spans to assist the wearer in performing a movement or maintaining a posture.

[0120] As an example, a representative passive wearable device 100 shown in FIG. 2A can be adapted to be a semi - active wearable device 100 by including a semi - active actuator 120 (e.g., a clutch system, a non - backdrivable actuator, or other suitable mechanism) configured to control the amount of tension generated by the wearable device 100 based on the wearer's motion or posture. The semi - active wearable device 100 extending from the shoulder to the thigh can be configured to control the load of the wearable device 100, and the wearer's motion or posture generates a moment around at least one of the following: (i) one or more hip joints of the wearer, and (ii) one or more spinal joints of the wearer. This is described in more detail throughout the present disclosure.

[0121] In various embodiments, the semi - active actuator 120 can be configured to control the load of the wearable device 100 by controlling parameters such as the following.

[0122] (a) Initial length of connecting element 150 : The initial length (x01) of the connection element 150 is defined as the time when tension begins to be generated as the user bends relative to the standing position, and the distance between both anchor points when the user is standing is defined as x0. For example, if the initial length (x01) of the anchor element is greater than or equal to x0 (x01 >= x0), tension is generated as the user bends the torso or legs for any activity relative to the standing position. This connection element 150 can be designed to be shorter than x0 (x01 < x0) if it is intended to provide tension while the user is standing. 4A and 4B show how the output tension can be changed by defining different initial lengths of the connection element 150. The process of adjusting can include commanding a target position of the actuator 120 and using a sensor 230, such as an encoder or potentiometer, to define the current position of the actuator 120 and close the loop. Once the initial length is defined, the actuator 120 can be clutched in that position so that it does not need to support the forces that arise in assisting the joint, or it does not need to be non-backdrivable, meaning that the motor does not need to use current to hold the position, which saves energy compared to a backdrivable system holding the position while the assisting forces are occurring.

[0123] (b) Slack vs. Pretension Mode : If the device is intended to be completely transparent (does not apply assistive forces), the actuator 120 system or clutch 160 can be used to allow slack in the system. In this case, the actuator 120 is commanded to release the cable 150 so that there is enough slack (the actuation cable 150 is longer than the distance between the anchor points), to unclutch the connection element 150 so that it cannot generate forces, or to open the pneumatic system so that the connection element 150 does not apply forces to the body. In one embodiment, the system can use a motion sensor 230, such as an IMU, to detect the user's posture and determine whether it is an activity that the device 100 can support or not. For example, by classifying activities such as walking, standing, lifting, holding a static posture, climbing stairs, etc., the algorithm defines which one it supports and commands the actuator 120 or clutch 160 to be transparent (does not generate forces), otherwise by adding slack to the connection element 150 between the anchor points.

[0124] (c) Load on connecting elementsThe clutch system 160 or actuator 120 may be used to select between various loads to add or remove load from the system by connecting various elastic elements, springs, or other passive connection elements 190 in parallel or by controlling the initial positions of various passive connection elements 190 in parallel.

[0125] (d) Compensation for movement of parts of the wearable device 100 In one embodiment, load cells on the connected elements 150 and motion sensors 230 (e.g., IMUs) can be used to monitor the forces resulting from a joint angle. If the resulting forces for a particular joint angle are less than desired, it is possible that the wearable device 100 is drifting, so that the elements 150 connecting the anchor points are not as pretensioned as they were when the wearable device 100 was first set up. A semi-active system can correct this pretension by removing slack from the system, following the methods described above. In another embodiment, a displacement sensor 230, such as a potentiometer, is used to measure the distance between the anchor points, in which case, for a given pose (e.g., defined as soon as the load cells measure zero force, or defined by a joint angle measurement, such as an IMU, as a reference), if the distance between the anchor points decreases, as the wearable device 100 moves over time, the actuator 120 may be controlled to remove slack from the system.

[0126] (e) Selective harvesting and delivery of wearer-generated energyIn one embodiment, the connection element 150 may include an energy storage device 190 (e.g., a spring) configured to absorb energy generated by a wearer's movement or posture, and the semi-active actuator 120 (e.g., clutch 160) may be configured to selectively lock the energy storage device 190 to store the absorbed energy. The semi-active actuator 120 may then be configured to unlock the energy storage device 190 to release the stored energy to assist the wearer in performing a movement or holding a posture. Although the released energy may actively generate moments about one or more of the wearer's joints, such an embodiment may still be classified as a semi-active wearable device 100 because the source of energy for the assisting forces and moments is the wearer's movement or posture, rather than a power source such as a battery. Active Wearable Devices

[0127] The wearable device 100 of the present disclosure, in various embodiments, may be configured to passively generate a moment about one or more joints of a wearer to assist the wearer in performing a motion or maintaining a posture. A passive wearable device 100 may generally include at least one anchor member 110 configured for placement on a first body part of a person wearing the wearable device 100, at least one anchor member 110 configured for placement on a second body part of the person wearing the wearable device, at least one connection element 150 that directly or indirectly couples the anchor member 110 of the at least one first body part to the anchor member 110 of the at least one second body part, and at least one actuator 120 configured to generate a tension force within the wearable device 100 to generate a moment about one or more joints spanned by the wearable device 100.

[0128] For example, the exemplary passive wearable device 100 shown in FIG. 2B includes two upper body anchor members 110 (e.g., shoulder anchor members 110 in the form of shoulder straps), two lower body anchor members 110 (e.g., thigh anchor members 110 in the form of thigh wraps), and connection elements 150 (e.g., actuation cables) connecting the upper body anchor members 110 to the lower body anchor members 110. As configured, the connection elements 150 can be configured to be actively actuated by the actuators 120 to generate a moment about at least one of: (i) one or more hip joints of the wearer, and (ii) one or more back joints of the wearer, as will be described in more detail throughout this disclosure.

[0129] As mentioned above, the wearable device 100 can be configured to assist different joints by creating anchor points on either side of the joint. An assist profile can be defined by a predefined timing and magnitude of actuation. In some embodiments, the actuation profile can be further reduced to a set of finite dimensional numerical variables representing key actuation parameters such as: (i) the timing of the start, peak and end of the actuation, (ii) the magnitude of the actuation, (iii) coefficients for polynomial interpolation of the actuation, (iv) a number of points that can be interpolated using mathematical functions such as splines or polynomials, and / or (v) the stiffness of the actuator 120.

[0130] The developed assist devices may be configured to support various joints, such as the back, thigh, knee, shoulder, elbow, or combinations thereof.

[0131] Referring back to FIG. 2B, a representative structure for generating assist torque around the hip and back joints is shown. This device can be completely transparent to other motions (like normal clothing) while serving to assist the back and hip joints during tasks such as lifting, standing still, carrying heavy equipment, i.e. pulling / pushing, which are common activities that can lead to injury over time. A representative device can be load controlled so that it can follow a predefined trajectory or provide a virtual load (stiffness, damping) that is a function of the joint angle. In this case, the load can be defined as the relative angle between the torso and the thigh, as the device spans both the hip and back joints. This angle can be measured by using an on-board sensor 230, such as an iMU, or a strain sensing element. A sample virtual load provided by actuation can be defined by a stiffness value, as shown in FIG. 4B, where F is the force produced, θ is the relative angle between the thigh and torso, and K is the stiffness provided. In this case, the assist profile feels like there is a spring between anchor points of stiffness K. The advantage of such actuation profiles over passive devices is that it allows more flexibility to automatically configure the assistance profile by selecting different loads optimized for different movements and different users. Furthermore, virtual loads have the advantage that they can have different properties (provided stiffness, damping, etc.) for each stage of the movement, for example different loads depending on the direction of the movement. Bending to grab an object requires less energy than reaching to grab an object while fighting gravity, and the change in the provided load when the movement changes direction can provide the person with more energy when extending versus bending, something that passive springs cannot do. Furthermore, by selecting different "initial angles", the final profile can be shifted so that no force is generated until the user reaches its initial position.

[0132] 5A, 5B and 5C show a sample user motion in which the joints follow a sinusoidal pattern, moving from an initial position of 0° (upright) to 90° (bending the torso to a horizontal position). This motion can be representative of an example of bending the torso to grab an object from the ground. The resulting force is the product of a given stiffness multiplied by the joint angle the user is performing. The force is expressed as a % of the lift cycle, where 0% is defined as when the user is initializing the motion, 50% is defined as when the joint velocity is zero as the user is changing direction, and 100% is defined as when the user has completed the motion by returning to an upright position.

[0133] In the case of the load control device 100, the system can have different modes depending on the detected motion. Thus, the on-board sensor 230 can be used to detect different activities and select different predefined load profiles for each of them. For example, there are multiple ways to lift an object, using, as examples, a squatting technique or a leaning forward (legs straight) technique. Also, the lifting may be done with the object in front of the body (back moving in the sagittal plane) or due to space constraints, with the object to the side and the back needing to bend or twist during the lifting. These different motions can be detected by using the on-board motion sensor 230, such as an IMU, that can define an initialization threshold (e.g., an initial threshold of the relative angle and angular velocity between the torso and thigh joints when passed) that is considered as the initialization of the bending motion when passed. A threshold can also be used to define whether the motion is occurring in the sagittal plane or if the user is bending or twisting the torso to lift. Another threshold can be used to define whether the user is bending or not bending the legs while lifting. In another example, the on-board sensors 230 can be used in a classification algorithm (e.g., neural networks, linear classifiers, nearest neighbors, decision trees, etc.) to detect which activity a movement is classified into. A load profile, or assistance profile, can be selected for each of these movements.

[0134] A load profile or assistance profile can be selected for each of these operations.

[0135] In another example, the device may be EMG controlled, where one or more EMGs are placed on representative muscles that the device is intended to assist. Filtered EMG signals from the muscles of interest may be used to estimate the joint torques that the joints of interest are generating, which may be used to command the delivered forces generated by the exoskeleton. Optimizing the control of wearable devices

[0136] Because these tasks are characterized as either having multiple iterations or being sustained for long periods of time, optimization methods can be used to measure and / or estimate the objective function over each of the multiple iterations or time and update the parameterized assistance profile to optimize the objective.

[0137] The optimization approach can be used to maximize or minimize the benefits of wearing a wearable device 100, such as an exoskeleton, outer garment, or robotic garment, during various types of activities, including non-locomotion and upper body movements. The wearable device 100 and optimization approach can assist different joints, such as the ankle, knee, or thigh during non-locomotion, or the elbow, neck, shoulder, or back during upper body movements, to enhance the performance (e.g., reduce fatigue) of healthy individuals or improve the efficiency of movement in impaired individuals.

[0138] As an example, Fig. 6 shows how two different healthy individuals obtain different results depending on the selected actuation parameters (in this case actuation timing), and how the maximum capacity of the system is reached with very different actuation parameters for each subject. This highlights the need to personalize the wearable device 100 (e.g., robotic garment, outer garment or exoskeleton) controller 240 for different individuals. An online optimization algorithm that uses wearable sensors 230 to measure or estimate an objective function and adapts different parameters of the assistance profile to maximize or minimize that objective while a particular user is performing a task allows for a higher degree of personalization of assistance.

[0139] In the context of a control system for a wearable device 100, including a robotic garment, outer garment, or exoskeleton, optimization techniques generally involve selecting an objective function, selecting a method for evaluating the objective function (e.g., by direct measurement, by calculation based on measurements of relevant information, or by proxy), selecting one or more operating parameters to optimize to maximize or minimize the objective function, and selecting and applying the optimization method.

[0140] The systems and methods of the present disclosure may seek to optimize one or more parameters associated with the actuation of a wearable device 100, such as a robotic garment, outer garment or exoskeleton, to maximize or minimize a selected objective function. Generally speaking, the robotic garment, outer garment or exoskeleton operates according to an actuation profile that defines the timing, speed, magnitude, and overall shape of actuation within a prescribed period. Often, the actuation profile is presented in the form of specified actuator 120 positions (e.g., cable positions when a Bowden cable connects a motor to a part of the robotic garment, outer garment or exoskeleton) over the prescribed period. For example, with respect to upper limb movements or non-ambulatory movements, the prescribed period may be a repetitive motion of the wearer. Of course, any suitable period may be defined by the biomechanics of the wearer and the part of the wearer's body that is being assisted by the wearable device 100, such as a robotic garment, outer garment or exoskeleton.

[0141] The actuation profile can be decomposed according to the predetermined timing and magnitude of actuation. In some embodiments, the actuation profile can be further reduced to a set of finite-dimensional numerical variables representing key actuation parameters such as: (i) the timing of the start of the actuation, its peak, and its end, (ii) the magnitude of the actuation, (iii) the coefficients of a polynomial interpolation of the actuation, (iv) a number of points that can be interpolated using mathematical functions such as splines or polynomials, and / or (v) the stiffness of the actuator 120.

[0142] Various types of actuators 120 may be used with the wearable device 100, including, but not limited to, pneumatic, hydraulic, electromechanical, electromechanical cable-based, electroactive materials, and actuators 120 that can generate force and / or change stiffness.

[0143] Various aspects of the wearer's biomechanics can be influenced and optimized using optimization techniques in conjunction with the garment or exoskeleton. For example, tasks that are cyclical in nature, involve multiple repetitions, or are sustained over long periods of time can utilize optimization algorithms to maximize or minimize an objective function that is evaluated during each repetition or over time, respectively. Exemplary cyclical upper body motions include repetitive work tasks, rehabilitation exercises, and assisted movements such as sit-to-stand movement and lifting loads, all common during daily activities.

[0144] Activities involving the upper body, including but not limited to the arms, neck, hands, wrists, shoulders, and / or back, can be enhanced by wearing an assistive wearable device 100, such as a robotic garment, outer garment, or exoskeleton, which can reduce fatigue, strain, and / or injury. For example, assistance can be provided to the back, shoulders, elbows, arms, and / or hands when lifting objects or performing overhead tasks. Also, a grip-assist device can increase the grip strength of the wrist and / or hand to grasp a tool or depress the tool's trigger, avoiding muscle strain and carpal tunnel symptoms if chronic. In all these cases, the wearable device 100 can be applied to the wearer while performing these tasks to reduce fatigue, improve ergonomics, prevent injury, or improve comfort. The wearable device 100 may be optimized and used on or across various joints of the wearer, including but not limited to the ankle, knee, thigh, back, neck, shoulder, elbow, hand, or a combination of the aforementioned joints. Because these tasks are characterized by multiple iterations or durations, various optimization methods may be used to measure and / or estimate objective functions and update parameterized assistance profiles over multiple iterations or over time to optimize the desired objective.

[0145] In the context of a garment or exoskeleton control system, optimization techniques may generally involve: (1) Selection of objective function (2) the choice of how to evaluate the objective function (e.g., by direct measurement, by calculation based on measurements of relevant information, or by proxy); (3) selecting one or more operating parameters to be optimized to maximize or minimize an objective function; (4) Selection and application of optimization methods

[0146] As will be described in more detail below, the present disclosure is directed, in relevant part, to a wearable device 100 comprising:

[0147] At least one actuator 120 configured to generate a force within the wearable device 100 or to cause a force to be generated within the wearable device 100, such that the wearable device 100 generates a moment about one or more joints of the wearer to assist the wearer in performing a movement or holding a stationary movement.

[0148] At least one sensor configured to measure information for evaluating an objective function relating to at least one of providing physical assistance to the wearer, an interaction between the wearer and the wearable device 100, and an operation of the wearable device 100.

[0149] At least one controller 240 configured to actuate at least one actuator 120 according to at least one actuation profile, evaluate the objective function based on information measured by the at least one sensor to determine a resulting change in the objective function, adjust at least one parameter of the at least one actuation profile based on the resulting change in the objective function, and continue to actuate, evaluate, and adjust to optimize the at least one actuation parameter to maximize or minimize the objective function.

[0150] As will be explained in more detail below, the present disclosure relates to a wearable device 100 comprising, in relevant part:

[0151] At least one actuator 120 configured to generate a force within the wearable device 100 or to cause a force to be generated within the wearable device 100, such that the wearable device 100 generates a moment about one or more joints of the wearer to assist the wearer in performing a movement or holding a stationary movement.

[0152] At least one sensor configured to measure information for evaluating an objective function relating to at least one of providing physical assistance to the wearer, an interaction between the wearer and the wearable device 100, and an operation of the wearable device 100.

[0153] At least one controller 240 configured to actuate at least one actuator 120 according to at least one actuation profile, evaluate the objective function based on information measured by the at least one sensor to determine a resulting change in the objective function, adjust at least one parameter of the at least one actuation profile based on the resulting change in the objective function, and continue to actuate, evaluate, and adjust to optimize the at least one actuation parameter to maximize or minimize the objective function.

[0154] As will be described in further detail, in another embodiment, the at least one controller 240 may be configured as follows.

[0155] identifying two or more operating parameters to be optimized to maximize or minimize said objective function;

[0156] actuating at least one actuator 120 according to two or more actuation profiles having different sets of baseline values ​​for the two or more actuation parameters to be optimized;

[0157] evaluating the objective function for each of the two or more actuation profiles based on information measured by the at least one sensor;

[0158] defining a mathematical correlation between the two or more operating parameters and the corresponding evaluations of the objective function based on the corresponding evaluations of the objective function;

[0159] evaluating the baseline mathematical correlation to determine a set of candidate values ​​for the two or more operating parameters that maximize or minimize the objective function;

[0160] updating a mathematical correlation based on a corresponding evaluation of the objective function for actuation of the at least one actuator 120 according to an actuation profile associated with a candidate set of values ​​for the two or more actuation parameters;

[0161] The mathematical correlation continues to be updated until the evaluation of the objective function reaches a global maximum or minimum, or until a predetermined termination criterion is met. Operating Profiles and Parameters

[0162] The systems and methods of the present disclosure seek to optimize one or more parameters associated with actuation of the garment or exoskeleton to maximize or minimize a selected objective function. Generally speaking, the exoskeleton or garment operates according to an actuation profile that defines the timing, speed, magnitude, load, and / or overall shape of actuation during a prescribed period. Often, the actuation profile is presented in the form of actuator 120 positions (e.g., cable positions when the Bowden cables connect a motor to a portion of the garment or exoskeleton).

[0163] Referring to FIG. 7, the actuation profile can also be based on force control as shown in FIG. 7 (right). Here, the prescribed period is the wearer's lifting cycle, which is a useful time frame when assisting the wearer during activities such as lifting. Of course, any suitable period or function can be specified depending on the wearer's biomechanics and the part of the wearer's body that is being assisted by the garment or exoskeleton. As an example, a system controlled in a loading mode (e.g., stiffness, damping) where the resulting force is a function of the wearer's motion may not require a period definition, but rather a function relating the desired force with respect to body posture or motion, as shown in FIG. 7 (left).

[0164] In Figure 7, the actuation profile can be decomposed according to the predetermined timing, rate, magnitude, and overall shape of the actuation. More specifically, the actuation profile can be further reduced to a set of finite-dimensional numerical variables that represent the key actuation parameters, such as: (i) the timing of the start of the actuation, its peak, and its end (Figure 7, right), (ii) the actuation rate, (iii) the magnitude of the actuation, and (iv) the coefficients of a function that defines the shape of the actuation, such as the coefficients of a polynomial, linear, exponential, sinusoidal, or other function that defines the shape of the actuation profile, such as impedance (e.g., stiffness, damping) as shown in Figure 7 (left).

[0165] Actuation of the garment or exoskeleton in combination with the wearer's natural movements causes the garment or exoskeleton to deliver forces and moments to the wearer's body. Similar to actuation profiles, the timing and magnitude of the resulting forces and moments may be characterized by a positional assist profile, as shown in FIG. 7. Here, actuation of the garment or exoskeleton according to the actuation profile of FIG. 7 (right) delivers forces and moments to the wearer's back as shown in the assist profile. In other cases, the assist profile may be directly parameterized in load space controlled by the force controller 240, as shown in FIG. 7 (left). In this example, the actuation is configured to assist the wearer's back joints when lifting an object. Objective functions and proxies, in general

[0166] A common goal of the optimization techniques disclosed herein is to maximize the benefits provided to the wearer by the robotic garment, outer garment, or exoskeleton. These benefits vary depending on a given application, but generally include providing physical assistance to the wearer. This may include, for example, assisting the wearer's natural movements (e.g., reducing energy expenditure, increasing strength), helping stabilize the wearer while stationary or in motion, or facilitating improvements in the wearer's periodic or repetitive movements (e.g., reducing muscle forces or improving posture for lifting tasks), among other suitable applications. Depending on the wearer and the objectives of the outer garment or exoskeleton, several metrics may be identified that, if maximized or minimized, can make the outer garment or exoskeleton more beneficial to the wearer. The optimization techniques disclosed herein seek to optimize one or more parameters, or objective functions, of how the outer garment or exoskeleton operates with the goal of maximizing or minimizing these metrics.

[0167] The systems and methods of the present disclosure, in various embodiments, utilize feedback from one or more sensors 230 of the wearable system to evaluate the impact of actuation adjustments on a given objective function. This feedback helps ensure that the wearable system converges on these actuation adjustments in a manner that maximizes or minimizes the desired function. Thus, in various embodiments, the controller 240 may be configured to evaluate the objective function in response to a given actuation of the garment or exoskeleton. As used herein, evaluating the objective function broadly includes any suitable method of quantifying the objective function. For example, in one embodiment, evaluating the objective function may be obtaining a direct measurement of the objective function from one or more sensors 230 included in the wearable system. In other embodiments, evaluating the objective function may include calculating or approximating the objective function using one or more measurements provided by one or more sensors 230 of the wearable system. For example, if the objective function is power delivered to the wearer by the garment or exoskeleton, the controller 240 may be configured to calculate power by multiplying measurements of force and rotational velocity measured by the load cell and IMU to calculate this criterion. In yet another embodiment, evaluating the objective function may include measuring, calculating, or approximating a proxy that correlates with the objective function using information measured by one or more sensors 230 of the wearable system to estimate the objective function via this correlation.

[0168] The use of proxies in evaluating the objective function may be particularly well suited for situations where it may be impossible or otherwise difficult to measure the objective function directly or to estimate it by calculation using relevant measurements from the wearable sensor 230. For example, it may be difficult to directly measure metabolic expenditure as a measure of exercise, typically requiring expensive equipment or requiring the user to wear a mask to measure gas exchange. However, other metrics, such as the ability to sustain motion for a longer period of time compared to a baseline, the ability to sustain a cyclical motion (e.g., lifting an object) at a normal frequency, or changes in posture may be used as objective metrics, as they may correlate with fatigue. As another example, delivering power to a joint during dynamic motion may reduce the level of effort for a particular task, so an optimization algorithm may use the positive force delivered to the organism's joints as a proxy for the level of joint force offloading. Of course, it should be noted that average positive force, changes in posture, along with any other proxies identified in this disclosure, may themselves be used as objective functions within the scope of the present framework. Stated another way, an objective function, as that term is used herein, does not necessarily represent an ultimate desired effect (e.g., maximizing locomotive efficiency), but can be any metric that is sought to be maximized or minimized (including within a set of range constraints) through optimization of one or more operating parameters.

[0169] As another example, heart rate has also been found to correlate with the wearer's level of exertion in some applications, and thus heart rate may be used as a proxy for force level in one embodiment. In yet another example, muscle activity of the primary muscles of the joint of interest may be correlated with force level and therefore used as an objective measurement.

[0170] Generally speaking, objective functions associated with robotic garments, garments or exoskeletons can be grouped into three categories: (i) those objective functions related to providing physical assistance to the wearer, (ii) those objective functions related to interactions between the wearer and the garment or exoskeleton, and (iii) those objective functions related to the operation of the garment or exoskeleton, as described in further detail below. Of course, these categories are presented for ease of explanation, and it will be understood that the optimization techniques described herein may be used in connection with optimizing actuation to maximize or minimize other suitable objective functions that fall outside of one of these three general categories. Objective functions and proxies related to providing physical assistance to the wearer

[0171] As mentioned above, many objective functions related to the capabilities of an outer garment or exoskeleton relate to the effectiveness of the wearable system in providing physical assistance to the wearer. Below, we provide representative examples of several such objective functions, along with biomechanical aspects of how each may be maximized or minimized to benefit the wearer.

[0172] Similarity between joint movement characteristics and normative data An exemplary optimization technique can assess how similar the wearer's motion characteristics are to age-, sex-, weight-, and / or speed-matched data from a normative set of healthy users or ergonomically recommended postures, and optimize actuation parameters to match the wearer's motion characteristics to the desired baseline profile (e.g., maximize similarity or minimize dissimilarity). Usable metrics include correlation with user data or percentage of difference in key points or metrics (e.g., torso angle during lifting, distance to COM of object). Similarity of joint motion characteristics and baseline dataAn exemplary optimization technique may evaluate how similar the wearer's motion characteristics are to baseline motion characteristics recorded when the user is wearing the system in transparent mode (without applying assistance) or not wearing the system at all, and optimize actuation parameters to match the wearer's motion characteristics to the baseline motion characteristics or posture (i.e., maximize similarity or minimize dissimilarity). Possible metrics include correlation with user data or percentage of difference in key points or metrics (e.g., torso angle during lifting, distance to COM of object).

[0173] Joint movement characteristics Exemplary optimization techniques may seek to reduce or minimize the biological torque generated by the joint of interest. Biological joints can be measured or estimated by combining kinematic measurements and ground reaction force measurements or estimates. Sample devices can incorporate pressure-sensing insoles to measure or estimate joint torque or power. Kinematic metrics that can be used include metrics that minimize peak torque or RMS joint torque, which minimize peak power or RMS power.

[0174] Time integral of muscle activity : This is the integral of muscle activity with respect to time. An exemplary optimization technique may seek to optimize actuation parameters to reduce or minimize the overall activity of the active muscles. As an example, one may wish to reduce or minimize the overall force of the back muscles when lifting or maintaining a stationary posture, or minimize the overall force of the shoulder muscles when performing overhead tasks.

[0175] Peak Muscle Activity : Similar to the time integral of muscle activity, but for peak muscle activity. This may be selected in some applications over or in combination with the time integral of muscle activity, for example, for activities where peak muscle activity may pose a risk of injury or where the peak muscle activity is greater than a person can perform over time (elderly applications, industrial workers, consumer applications).

[0176] Similarity of muscle activity profile to baseline or normative data An exemplary optimization technique may evaluate how similar the wearer's muscle activity is to baseline measurements recorded when the user is wearing the system in transparent mode (without applying assistance), or when not wearing the system at all, or based on normative data for a particular activity, and optimize actuation parameters to match the wearer's movement characteristics to the baseline or normative muscle activity (i.e., maximize the similarity or minimize the dissimilarity).

[0177] Similarity of muscle activity in muscles not assisted by the device An exemplary optimization technique may evaluate how similar the wearer's muscle activity (e.g., RMS and / or peak) is to baseline measurements recorded when the user is wearing the system in transparent mode (without applying assistance), or when not wearing the system at all, or based on normative data for a particular activity, and optimize actuation parameters to match (i.e., maximize similarity or minimize dissimilarity) the actuation to the baseline or normative muscle activity. As an example, while lifting with a device that supports the back muscles, the controller 240 may measure or estimate muscle activity of the abdominal muscles to avoid undesired interference with muscle groups not supported by the device.

[0178] Ergonomic Metrics and IndicatorsThe field of ergonomics in industry has defined multiple methods to estimate the risk of injury when performing a task. The controller 240 can use these equations or indices to minimize the risk of injury while assisting a robotic garment, outer garment, or exoskeleton. As an example, NIOSH defines lifting to evaluate how significant the risk of performing a lift is. The equation is a function of the horizontal and vertical distance of the body to the object, the asymmetric angle when lifting the object, and the frequency and duration of each activity among other factors. The resulting index is used as part of the objective to minimize the risk of injury, and any change in the movement / dynamic characteristics compared to the baseline will result in a change in the lifting index. An algorithm evaluates this index while the person is performing the task and minimizes it by comparing it to the baseline or normative data.

[0179] Energy consumption : quantifies the level of energy expenditure or force when performing a task. Systems that measure this variable may use it as an objective function in algorithms to reduce metabolic effort for certain tasks (e.g., lifting an object, holding a static position, pushing, lifting, grasping, etc.). This metric is useful for a variety of applications, including healthy individuals, individuals with disabilities, and the elderly performing strenuous tasks. This objective function is also referred to herein as metabolic effort.

[0180] Stability during lifting Exemplary optimization techniques may seek to optimize actuation parameters to maximize joint stability or overall body stability while lifting or holding a static position. Stability metrics can be used as part of the optimization method to improve stability during lifting and other tasks for healthy individuals performing complex or strenuous tasks.

[0181] Range of motion: An exemplary optimization technique may seek to optimize actuation parameters to maximize the range of motion of a particular joint. Range of motion is typically defined as the difference between the maximum and minimum angles of a particular joint in a particular plane of motion, although maximum and minimum angles may also be useful for optimization. For example, the range of motion of the trunk or thigh in the sagittal or dorsal plane during a task such as lifting.

[0182] Time to perform repetitive tasks : An exemplary optimization technique may seek to optimize operating parameters to minimize the time required to perform a repetitive task, such as lifting a load or moving a package from one location to another in a factory environment.

[0183] Similarity of foot center of pressure location to normative data An exemplary optimization technique may seek to optimize actuation parameters to maximize the similarity (or minimize the dissimilarity) between the trajectory of the wearer's center of pressure during lifting and a desired baseline or normative trajectory. The center of pressure can be measured using a force plate or estimated with a wearable sensor 230, such as an instrumented shoe insole.

[0184] Blood oxygenation in muscles Related to pulse oximetry commonly used in clinical practice, near-infrared spectroscopy (NIRS) is a method to estimate how well blood in a muscle is oxygenated. Because this measurement is directly related to the metabolic demand of that muscle, this metric can be used as a proxy to evaluate relevant objective functions (e.g., energy consumption, locomotive efficiency). An exemplary optimization technique might seek to optimize operating parameters to maximize blood oxygenation, for example, in situations where it is desired to minimize energy consumption or maximize efficiency.

[0185] Muscle contraction / tension and tendon tensionWearable garments and exoskeletons can affect how muscles contract and muscles and tendons tense under load. Exemplary optimization techniques may seek to optimize actuation parameters to maximize or minimize these physiological parameters, or to limit them to specific values ​​or ranges. Furthermore, since contraction / tension is often related to the energy metabolic demands of the corresponding muscles, this metric may be an appropriate proxy for evaluating relevant objective functions (e.g., energy consumption, locomotive efficiency, etc.). Exemplary optimization techniques may seek to optimize actuation parameters to minimize muscle contraction / tension and tendon tension, for example, in situations where it is desired to minimize energy consumption or maximize efficiency. Ultrasonography or sonography may be used in some embodiments as a method of non-invasively measuring tissue movement.

[0186] Exemplary optimization techniques may utilize information relevant to providing physical assistance to the wearer in evaluating these objective functions. For example, in some embodiments, controller 240 may evaluate these objective functions using direct measurements of the objective functions themselves, while in other embodiments, controller 240 may evaluate these objective functions by calculating or approximating the objective functions using various measurements of relevant metrics from one or more wearable sensors 230 or other measurement devices. In various embodiments, representative information relevant to these objective functions may include any of the information described above, as well as one or a combination of the following: angles, velocities, or accelerations of one or more joints, forces, torques, or powers applied by the robotic garment, outer garment, or exoskeleton, the type of physical activity the wearer engages in, the wearer's posture, the wearer's energy expenditure, and the interaction of the wearer's body with the surrounding physical environment.

[0187] Exemplary sensors 230 for measuring the information for evaluating the objective function may include, in various embodiments, inertial measurement units (IMUs), accelerometers, gyroscopes, encoders, resolvers, strain sensors 230, heart rate monitors, pulmonary gas exchange systems, electromyography, pressure sensors 230 (e.g., shoe insoles to measure GRF), and other suitable sensors 230. Objective functions and proxies related to the interaction between a wearable device and its wearer

[0188] As referenced above, many objective functions related to the capabilities of the wearable device 100 relate to the interaction between the outer garment, exoskeleton, or robotic garment and the wearer. As explained further below, such interactions may relate to the comfort of the wearer, the effectiveness of the wearable system in transferring loads to the wearer, etc. Representative examples of several such objective functions are provided below, along with biomechanical aspects of how each may be maximized or minimized to benefit the wearer.

[0189] Displacement of Wearable Devices 100: An exemplary optimization technique may optimize actuation parameters to attempt to minimize (or maximize adjustment of) the displacement of the wearable device 100 (including its individual parts, such as the anchor member 110) on the wearer's body. The displacement of the parts of the wearable device 100 relative to the body may be measured by using a wearable sensor 230, such as an IMU. As an example, by using two IMUs, one integrated in the parts of the wearable device 100 and the other sensor attached to the body, to integrate measurements such as relative velocity and acceleration. For cable-driven systems with cables fixed on both sides of the joint, the relative displacement may be estimated by combining the exposed cable (e.g., by an encoder) and the measurement of the joint position using a wearable sensor 230, such as an IMU. The amount of exposed cable is a function of the cable required to track the joint motion, the stiffness of the wearable device 100, and the compliance of the joint parts. If displacement is measured at a point in the gait cycle where no active forces are being applied to the person (only pretension or tracking with no applied forces), the stiffness of the wearable device 100 does not affect the movement of the cables, and thus displacement can be calculated by combining the movement of the cables through the joint angle (converted to displacement by multiplying the joint's center of rotation and distance from the cable, which may be a constant or an input to the device) and the sum of the measurements of the exposed cables. Finally, for cable-driven systems where the cables are fixed on either side of the joints, the motor's encoders may be used to estimate relative displacement by comparing the exposed cables to a repeatable point over multiple repetitions, such as a standing posture, which gives an estimate of the amount that parts of the wearable device 100 are displaced relative to the body over time.

[0190] ComfortExemplary optimization techniques may attempt to optimize operating parameters to maximize comfort for the wearer (or minimize certain discomforts experienced by the wearer). In various embodiments, potential discomfort may be detected automatically by the controller 240, for example, by monitoring the forces at the wearable device 100 or the pressure between the wearable device 100 and the wearer and comparing these forces and / or pressures to respective levels known to cause discomfort for an individual wearer or across a representative sampling of wearers. If the wearable device 100 is, for example, generating forces that exceed comfortable levels (or at inappropriate times that cause discomfort), an exemplary optimization technique may attempt to optimize operating parameters to reduce these forces to within acceptable limits (perhaps while still maximizing the levels of force generated within acceptable limits). Similarly, if the interaction between the garment or exoskeleton and the wearer generates excessive levels of shear forces (which may cause chafing or other discomfort on the skin), a representative optimization approach may seek to optimize actuation parameters to reduce these shear forces to within acceptable limits (perhaps while still maximizing the level of forces generated within acceptable limits). In practice, this may involve reducing the level of forces generated by the wearable device 100, or perhaps increasing compressive forces within the wearable device 100 to better secure the anchor member 110 in question to the underlying body part. As described further below, a similar approach may be employed to optimize actuation parameters to reduce pressure points generated by the wearer's interaction with the wearable device 100. Additionally or alternatively, comfort may be self-reported via a GUI, a handheld device, voice, pressure measurements, etc.

[0191] Pressure at the jointAn exemplary optimization technique may optimize actuation parameters to attempt to minimize pressure applied to some areas of the wearer's body that may be sensitive to discomfort caused by the wearable device 100 and fiber or rigid junctions. This may maximize comfort (or minimize discomfort) while maximizing assistance within these comfort levels, for example. This pressure may be affected by actuation parameters such as timing, speed, awkwardness of profile, or magnitude of assistance.

[0192] Load transfer along the wearer's body In some cases, the wearable device 100 may be utilized to transfer load from one area of ​​the wearer's body to another, for example, to help mitigate injury. An exemplary optimization technique may seek to optimize actuation parameters to maximize load transfer from the area of ​​the body in question. This may be particularly useful in situations where the wearer is lifting heavy loads and is susceptible to back injury. In such an embodiment, the controller 240 may seek to optimize actuation parameters to minimize load on the wearer's spine. This may be accomplished, in one embodiment, by optimizing actuation parameters to maximize force transfer to other areas of the body, such as the thighs, torso, or legs.

[0193] Actuator 120 force An exemplary optimization technique monitors interactions with the wearable device 100 and, upon detecting large forces generated by the interaction, optimizes actuation parameters to minimize the forces generated by the actuator 120 and avoid delivering additional forces that could cause discomfort or injury. In one embodiment, the controller 240 may keep other biomechanical constraints constant while optimizing the actuation parameters.

[0194] Exemplary optimization techniques may evaluate these objective functions utilizing information related to the wearer's interaction with the wearable device 100. For example, in some embodiments, the controller 240 may evaluate these objective functions using direct measurements of the objective functions themselves, while in other embodiments, the controller 240 may evaluate these objective functions by calculating or approximating the objective functions using various measurements of relevant metrics from one or more wearable sensors 230 or other measurement devices. In various embodiments, representative information related to these objective functions may include information such as those described above, as well as one or a combination of shear or compressive forces occurring between the wearer and the garment or exoskeleton, and the position of the garment or exoskeleton on the wearer.

[0195] Exemplary sensors 230 for measuring such information to evaluate the objective function may include load cells, force sensors, torque sensors, and pressure sensors, as well as other suitable sensors, in various embodiments. Objective functions and proxies related to the operation of wearable devices

[0196] As referenced above, many objective functions related to the capabilities of the wearable device 100 relate to the operation of the wearable device 100. As explained in more detail below, such interactions may relate to power consumption, overheating, consistency of actuator / actuation system functionality, etc. Representative examples of such objective functions are provided below.

[0197] power consumption An exemplary optimization technique may optimize operating parameters to minimize power consumption. Power consumption may be measured using a combination of current consumption measurements (via an internal current sensor) and voltage (via an internal voltage sensor), which may be multiplied together to calculate power, with the average power consumption over time providing an estimate of the amount of power consumed in a given activity. In other embodiments, the remaining battery voltage may be used to estimate the amount of battery remaining in the system or the amount of battery that has been consumed so far.

[0198] Consistency of support In an exemplary optimization approach, the operating parameters may be optimized to maximize the consistency (or minimize the difference) between the assistive forces generated by the wearable device 100. If the wearable device 100 is intended to provide a consistent force in terms of peak force over multiple iterations, the consistency may be measured by the variance of the peak force or other features in the force profile that need to remain consistent over multiple iterations. The force may be measured using a wearable sensor 230, such as a force sensor or load cell. In another embodiment, the consistency of the assistive force may be defined as the mean square deviation of the measured force and the desired force profile, and maximizing the consistency of this difference over multiple iterations may be used as an objective of these systems.

[0199] Maximum and average values ​​of actuator position, velocity, acceleration, or jerk Due to the kinematic or kinetic constraints measured as described above, it may be important to minimize these metrics, which relate to the level of force at which the wearable device 100 is moving. For example, an exemplary optimization technique may find an optimal parameter or set of parameters that minimizes the maximum cable acceleration without affecting the lift index.

[0200] temperature An exemplary optimization technique may seek to optimize operating parameters to minimize the temperature of one or more components of the wearable device 100. As with other objective functions described herein, minimizing the temperature means, in one embodiment, preventing the temperature from exceeding an upper threshold of a pre-defined range.

[0201] Exemplary optimization techniques may evaluate these objective functions utilizing information related to the operation of the wearable device 100. For example, in some embodiments, the controller 240 may evaluate these objective functions using direct measurements of the objective functions themselves, while in other embodiments, the controller 240 may evaluate these objective functions by calculating or approximating the objective functions using various measurements of relevant metrics from one or more wearable sensors 230 or other measurement devices. In various embodiments, representative information related to these objective functions may include, in addition to the information described above, one or a combination of the following: current or voltage used during actuation of the garment or exoskeleton, temperature of one or more components of the garment or exoskeleton, force generated by at least one actuator 120, force delivered to the garment or exoskeleton, and force, torque, or power delivered to the wearer of the garment or exoskeleton.

[0202] Exemplary sensors 230 for measuring the information to evaluate the objective function may include, in various embodiments, current sensors, voltage sensors, thermistors, encoders, and resolvers (e.g., for measuring actuator position, velocity, and acceleration), as well as other suitable sensors. Weighted Objective Functions and Proxies

[0203] As mentioned in several examples above, the objective function (or proxy function) does not have to represent a single variable associated with the wearer's motion or biomechanics, but can combine multiple optimization objectives. Multiple single-variable objective functions can be combined into one through various methods, such as weighted averaging, which allows adjusting the relative importance of each component of the objective function. Equation 1 below shows an example of a weighted sum of multiple objective functions. F = α1x1+α2x2+...αnxn, where

number

[0204] This approach allows us to pursue the combination of optimizing variables related to the physical assistance to the wearer (e.g., peak muscle activity during lifting) with variables related to the efficiency of the system (e.g., battery life). An example of weighting two or more objectives within one objective function is shown below:

[0205] Optimize variables related to the movement (the kinematic characteristics of the lift) and variables related to the device capabilities (e.g., the device's battery life).

[0206] Peak muscle activity, posture during the activity, device battery life, and comfort are all weighted together to optimize the result.

[0207] A number of variables related to movement (e.g., difference in movement characteristics compared to a baseline) and variables related to the wearer's interaction with the wearable device 100 (e.g., user comfort, etc.) are optimized.

[0208] The device optimizes peak muscle activity in targeted muscles without affecting muscle activity in key muscles that are not being supported.

[0209] Optimize multiple variables related to performance (e.g., stability and speed at which a task is performed).

[0210] Optimize guidelines that assess the risk of injury when performing a task (e.g. lift value) and the time to perform the activity.

[0211] Optimize the time to complete the task and the peak EMG activity of the targeted muscles.

[0212] The importance of each criterion may vary depending on the application. An example of this concept is shown in Figure 8 using a weighted sum of objectives for a disabled person. In able-bodied people, individual users may manually change these settings. Optimization methods, in general

[0213] The present disclosure provides systems and methods for optimizing one or a combination of these operating parameters to maximize or minimize a given objective function. Approaches for optimizing a single operating parameter (e.g., finding the optimal actuation start timing) are referred to herein as single-parameter optimization, and approaches for optimizing a combination of operating parameters (e.g., finding the optimal combination of actuation start timing and actuation peak timing) are referred to herein as multi-parameter optimization. As described further below, many factors can affect whether to use a single-parameter or multi-parameter optimization approach, including, but not limited to, the biomechanics of the wearer, the people wearing the wearable device 100, the relative impact of each operating parameter on the objective function, the required optimization time, measurement noise, wearer adaptability, and available computing power, as described further below.

[0214] In various embodiments, known (or modified) methods for optimizing variables (e.g., gradient descent, Bayesian, brute force, and simulated annealing) may be used or modified to optimize the target operating parameters to maximize or minimize a selected objective function. Of course, the best mathematical method for optimizing the operating parameters may vary depending on many factors, including but not limited to the nature of the situation for the objective function associated with the operating parameters being optimized, the biomechanics of the wearer, the people wearing the wearable device 100, the relative impact of each operating parameter on the objective function, the required optimization time, measurement noise, wearer adaptability, and available computing power, as further described below.

[0215] In one embodiment, a simplified gradient descent method may be used to optimize the operating parameters to maximize or minimize the objective function. Gradient descent methods generally lead to convergence and are often simple and computationally inexpensive to run on inexpensive processors 250 or on shared hardware for other computationally intensive processes. In one such embodiment, controller 240 may be configured to determine whether to increase or decrease the objective function from one or an average of prior evaluations of the objective function, and in response, may adjust the operating parameters in the same or different directions depending on whether the objective function is maximized or minimized. Similarly, if multiple actuation profiles are tested before adjustments are made, the controller 240 may be configured to compare the evaluation of the objective function for each actuation profile to determine which produces the highest or lowest evaluation of the objective function, and may correspondingly generate another set of actuation profiles in which the values ​​of the actuation parameters to be optimized are adjusted according to the derivative of the evaluation of the objective function associated with the previous set of actuation profiles. The gradient descent method may optimize the actuation parameters to a local maximum / minimum or a global maximum / minimum depending on the objective function context (i.e., the relationship between the objective function and the parameters to be optimized). For example, convergence to a global solution may be achieved if the objective function context is substantially convex in shape and the differentiation of the context is differentiable and bounded. Since the objective function context is convex, all local minima are global minima, so in this case the gradient descent method may converge to a global solution. The gradient descent method may optimize the actuation parameters to a global maximum / minimum of the objective function. In other situations, such as when the objective function landscape contains multiple peaks or valleys, gradient descent may optimize the operating parameters to either a local maximum / minimum or a global maximum / minimum, depending on relevant factors such as where to start the optimization. In situations where the landscape is not convex, gradient descent performs a local search that compares points that are relatively close to each other, and is therefore more likely to find a local optimum, whereas other approaches that perform a global search of the landscape, such as the Bayesian method described below, are more likely to find a global optimum in such cases. However, the low computational requirements of gradient descent and guaranteed convergence may make this method attractive, even in some cases, if limited. As an example, finding the location of a local optimum that improves the initial setup of a device for a particular person or activity may be attractive in some applications. Gradient descent is particularly applicable in situations where the noise-to-signal ratio is very low and the objective is measured in a reproducible way, and where the measurements are noisy even in convex landscapes, and gradient descent may converge to a point in the landscape that is not optimal due to noise or non-reproducible measurements.

[0216] In other embodiments, a Bayesian method may be used to optimize the operating parameters to maximize or minimize the objective function. Bayesian approaches are typically sample-efficient and can explore the objective function landscape relatively quickly. This feature may make Bayesian methods well-suited to find and optimize the operating parameters to global maxima / minima in more complex objective function landscapes, and Bayesian methods tend to be relatively more tolerant to noise in the evaluation of the objective function than some other methods. In one such embodiment, a mathematical correlation between the operating parameters and the objective function may be defined by operating the wearable device 100 according to a number of operating profiles having different baseline values ​​of the operating parameters to be optimized and mapping the corresponding evaluations of the objective function for each. This may allow the controller 240 to determine candidate values ​​of the operating parameters to maximize or minimize the objective function. The controller 240 may test and evaluate the candidate values ​​and update the mathematical correlation until convergence to an optimized value of the operating parameter or until some other suitable termination criterion is met. In one embodiment, the process of defining the mathematical correlation and identifying candidate values ​​for the operating parameters includes generating a posterior distribution (e.g., via a Gaussian process) and using a probabilistic model (e.g., in one embodiment, a Bayesian method) to identify candidate values ​​for the operating parameters based on the posterior distribution. Like gradient descent, depending on the objective function context and related factors such as where to start the optimization, a Bayesian method may optimize the operating parameters to a local maximum / minimum or a global maximum / minimum. However, because a Bayesian method performs a global search of the context based on the posterior distribution, as opposed to just comparing points that are relatively close to each other locally, as gradient descent does, it is more likely to find a global optimum. Thus, Bayesian optimization is particularly useful in situations where the objective function context is complex, where there are multiple parameters to optimize and search efficiency is important, or where the objective function measurements are noisy (e.g., metabolic measurements).

[0217] Of course, it should be recognized that these are merely illustrative examples and that any suitable mathematical technique may be used to optimize any number of operating parameters to maximize or minimize an objective function that is sensitive to the selected parameters.

[0218] In various embodiments, the optimization algorithm may be configured to terminate when one of the following predefined termination conditions occurs: (i) the difference in the evaluations of the objective function for continuously adjusting at least one parameter of at least one actuation profile falls below a predefined threshold, (ii) the difference in the continuous adjustments to at least one parameter of at least one actuation profile falls below a predefined threshold, (iii) a predefined amount of time has been exceeded since the start of the optimization, (iv) the wearer has exceeded a predefined number of iterations since the start of the optimization, or (v) the wearer issues an exit command.

[0219] As discussed above, various exemplary parameters related to the biomechanics of the wearer include muscle activity (e.g., time integral or peak muscle activity) of target muscles. For example, the parameters may be related to minimizing muscle activity of a particular task to reduce fatigue of the particular muscle when performing the task. Other exemplary parameters include, but are not limited to, minimizing the time to complete a repetitive task, joint kinematics or dynamics, joint range of motion, metabolic effort, and maximizing stability when performing the task. Objective functions related to the interaction between the robotic garment, the wearable device 100, the wearable device 100, and the wearer include, but are not limited to, load transfer, pressure at the interface, and comfort.

[0220] In some embodiments, sensors 230 configured to measure human biomechanics may include, but are not limited to, inertial measurement units (IMUs), accelerometers, gyroscopes, encoders, resolvers, and strain sensors 230 to measure joint angles, velocities, and accelerations, heart rate monitors to measure heart rate, pulmonary gas exchange systems to measure metabolic activity, electromyography, and pressure-sensing insoles and instrumented treadmills to measure ground reaction forces. Sensors 230 that can be used to measure wearer-system interactions, including forces and pressures, include, but are not limited to, load cells, force sensors, and pressure sensors. Sensors 230 that can be used to measure device metrics include, but are not limited to, current sensors, voltage sensors, thermistors, and encoders and resolvers to measure motor position, velocity, and acceleration.

[0221] FIG. 8 illustrates a representative approach for optimization control of a wearable device. Detecting motion initialization: In some applications, an optimization algorithm needs to be able to detect when an activity begins and when it ends, and evaluate the objective function using data collected during the iterations. As an example, for a device that supports the back during a lifting motion, the optimization algorithm may need to segment the lifting motion into different stages, such as motion initialization, moving to reach the object, grasping the object, and moving to a final posture until the lifting motion ends. The optimization algorithm may evaluate the objective function for each of these stages. As an example, the optimization algorithm may use an on-board motion sensor 230, such as an inertial measurement unit, to detect, for example, changes in joint angle, velocity, or acceleration, to segment the motion and evaluate the objective function. The algorithm then detects changes in joint angle, velocity, or acceleration, detects motion initialization, and starts evaluating the objective function, and measures or estimates the objective measure, for example, by seeing when joint acceleration changes direction or becomes equal to or close to zero (below a threshold) until the person ends the motion. Finally, if the acceleration or velocity changes, the objective function is evaluated as the user progresses to the final posture until the user stops the motion, which is recorded as one of the motion iterations by the optimization algorithm, which evaluates the objective function and determines the next set of parameters to be tested as optimal assistance candidates for the next motion iteration.

[0222] As an example, a control algorithm may consider a subgroup of different activities and find a set of parameters that maximizes or minimizes an objective function for each of those activities.

[0223] The various essential movements that the device assists can be categorized into subgroups, for example for a device assisting the back joint, these various subgroups could be lifting an object in front while keeping the knees straight (bending), lifting an object in front while bending the knees (squatting), lifting an object to both sides while bending or squatting followed by a twisting movement of the torso (a common movement in multiple industries such as construction, delivery, logistics, military logistics, manufacturing, etc.), and holding static positions with various torso angles (a common position for surgeons, caregivers, manufacturing, etc. that can put strain on the back and ultimately lead to injury).

[0224] For example, referring to (FIG. 9), an IMU may be used to detect when a movement begins (change in joint acceleration or velocity) and the direction of movement, e.g. bending and / or twisting of the torso, to define whether the user is performing a movement in the sagittal plane (e.g. lifting an object in front of the user) or bending and twisting of the torso (e.g. lifting an object sideways), which may be used to classify the movements described above. Additionally, synchronized movements of different body parts may be used to detect whether a person is bending up (legs are straight but torso is bending) or squatting up (bending of torso and legs).

[0225] The optimization algorithm then evaluates the objectives while the user performs the action and optimizes the set of parameters.

[0226] As the user repeats an activity that includes actions in the above subgroups, the device's parameters will be automatically modified to maximize or minimize the objective function, resulting in a set of parameters for each subgroup.

[0227] When a user performs an action, the control algorithm may detect what type of action the user is performing and select an appropriate set of parameters based on the subgroup to which the action belongs. Representative examples

[0228] For further context, various non-limiting examples of the wearable device 100 of the present disclosure are provided below. Representative Example 1A: Passive Back and Hip Support Device

[0229] As mentioned earlier, overexertion can often result in fatigue and musculoskeletal injuries. Back accounts for the majority of all overexertion injuries in industry. Risk factors that can lead to back injuries include heavy lifting, repetitive lifting, maintaining the same position for extended periods of time, or lifting heavy weights with suboptimal body postures.

[0230] Bending over while working is necessary in some work environments; however, this position can cause lower back strain, which over time can lead to chronic back pain or more serious back injury.

[0231] Some examples of activities that can benefit from a back support device include:

[0232] Lifting - Lifting heavy objects or lifting objects in non-ergonomic, non-optimal positions are activities that may require back support.

[0233] b. Holding or moving an object can require a lot of effort, potentially resulting in injury, and can be uncomfortable when performing a task, especially if the object is heavy or if the body position is not optimal, and therefore can benefit from a back support device.

[0234] c. Lifting and repositioning patients – In hospitals, healthcare workers may need to handle or move patients who need assistance. These types of activities are common and can benefit from back support.

[0235] d. Constant posture - for example, standing, crouching, or leaning forward, maintaining one position, if this position is maintained for a long period of time.

[0236] A robotic apparel system could be used to reduce the amount of effort that back muscles must exert during tasks such as lifting and maintaining a certain posture. Figure 10 shows an example of robotic apparel, where a support member 110 spans multiple joints, and the system simultaneously assists the back and hip joints during tasks such as lifting, holding heavy loads, and maintaining a certain posture for extended periods of time. During these tasks, both the hip and back joints are active, and therefore a device that assists both joints would reduce the amount of effort a user must exert.

[0237] This device may include the following components:

[0238] Sensor 230The system may include sensors 230 that can measure or estimate the velocity, acceleration, or position of different body parts, allowing the sensors 230 to estimate the amount of pulling force the device is providing to the user. Motion sensors 230, such as inertial measurement units, gyroscopes, accelerometers, etc., may be used for this purpose. Other examples of sensors 230 include pull sensors 230 that deform as the user moves and this deformation can be correlated to the acceleration, velocity, or position of the body part. In this embodiment, the system includes four inertial measurement unit sensors 230, one on each thigh to measure the thigh movement, one on the lower back, and one on the upper back, as shown in FIG. 9. These sensors 230 may be used to measure the movement of these body parts, but also to calculate the angle of the joint by subtracting the values ​​of the sensors 230 on either side of the joint. As an alternative to inertial measurement units, pull sensors may be used to correlate with the movement of the joint. The system may incorporate force sensing elements, such as load cells, force sensors, strain gauges, etc., to measure or estimate the pulling force the system is applying to the body. In this embodiment, these sensors 230 may be integrated into the apparel portion.

[0239] Wearable Device ConfigurationThe configuration of the wearable device 100 provides two anchor points on either side of the back joint. The upper part of the device is attached to the body by anchor members 110, such as shoulder straps, and the point where both shoulder straps meet at the user's back creates an anchor point located on one side of the back joint. The wearable device 100 may also include a lower body support member 110, such as a thigh support member 110, which wraps around to connect to the thigh and creates an anchor point on one side of the hip joint. A load balance strap 112 connects both thigh anchor members 110 and creates an anchor point on the other side of the back joint. Finally, a passive element 190, in this case an elastic fabric, connects between both anchor points and provides pulling forces to assist the user. Because these forces are located away from the joints that are the center of rotation, the forces provide a torque equal to the product of the pulling force and the moment arm. As described above, pulling forces between two joints, namely the hip joint and the lower back joint, provide assistance to the user. Pulling forces generate torques around the hip and lower back joints, as shown in the simplified diagram in Figure 10. For example, when a user lifts a load, the wearable device 100 can reduce the amount of effort his / her hip and back must exert. This means that biological muscles have to exert less effort, reducing fatigue and reducing the risk of injury to the wearer.

[0240] The proposed device has multiple applications for various industries. Figures 11 and 12 illustrate some representative applications.

[0241] One important requirement for a passive device is to provide forces or torques that are beneficial to the user when performing tasks such as lifting, maintaining a posture, etc. However, the system must be sufficiently transparent and not provide forces that would interfere with other movements such as walking, climbing stairs, etc. A load balancing element 212 serves this purpose.

[0242] Load Balance Strap 212 As described above, the load balance strap 212 connects both thigh anchor members 110 to create an anchor point on one side of the dorsal point. The load balance element 212 (e.g., strap, ribbon, cable) may be rotatably or slidably coupled to the load balancing mechanism 215, allowing the load balance strap 212 to rotate or slide relative to the linking element 150 during wearer movement, such as walking or lifting. Representative examples of the load balance mechanism 215 include, but are not limited to, low friction rollers, bearings, pulleys, low friction clasps, etc., as shown in FIG. 13. The load balance element 212 can rotate and slide relative to the load balance mechanism 215, allowing the user to move the legs freely when performing tasks such as walking, and creating tension in the wearable device 100 during tasks that involve bending the torso and / or both legs (lifting, picking up, etc.). Additionally, the load balancing element 212 distributes the pulling forces between both feet by splitting the load between both femoral anchor members 110 .

[0243] Figure 14 summarises the function of the load balance strap 112. When the user is standing, the initial distance between both support points is defined as x0. The elastic support element may be dimensioned or adjusted with a length equal to x0 such that it exerts no pulling force on the user while standing, but as soon as the user bends his / her torso and / or legs, it exerts a pulling force and generates tension.

[0244] As illustrated in the figures, if the user is walking, the self-balancing strap 112 can rotate freely relative to the anchor point, so that both legs can move freely without changing the distance between both anchor points (and therefore without generating a pulling force). If this element 112 could not rotate, moving the leg forward would generate a pulling force on the leg moving forward of the user and on the back of the user, which is undesirable. As shown in FIG. 13, when the wearer takes a step forward with the right leg while walking, the strap 112 can rotate relative to the anchor point, and therefore translates towards the leg moving forward without reducing the amount of tension created in the leg or generating tension, so that the leg can swing freely forward. In the intermediate position, while both legs are straight, the load balancing element 212 reaches a middle position, where the distance between the anchor point and both thigh anchor members 110 is approximately equal or relaxed (neutral position). As the person swings their left leg forward, the balance strap 112 rotates about the attachment point during the swing, allowing the left leg to move freely forward beyond the neutral position.

[0245] When a user bends his / her torso or legs, such as when picking up or lifting an object, the relative distance between both support points increases as a result of the user's kinematics. If this distance is longer than the initial dimension of the linking element 150, it causes the elastic element to stretch (x=x0+Ax), thus providing tension forces behind the user's back. These forces are proportional to the amount of stretching, due to the spring constant of the elastic body (F=k*Ax).

[0246] The sizing and selection of the passive coupling element 150 is based on the assistance desired: different jobs and individuals may require or prefer different types or degrees of assistance. Assistance may be tailored by determining one or a combination of the following parameters:

[0247] a) Initial length of the link element: The initial length of the passive connecting element (x01) is defined as the point when the tensile force starts as the user bends relative to the standing position, and the distance between both support points when the user is standing is defined as x0. For example, if the initial length (x01) of the support element is equal to or longer than x0 (x01 >= x0), as the user bends relative to the standing position of the torso or legs to perform activities, a tensile force is generated. If it is intended to supply a tensile force when the user is standing, this connecting element 150 may be designed to be shorter than x0 (x01 < x0). FIGS. 15 and 16 show how the output tensile force can be modified by defining different initial lengths of the connecting element 190. The procedure for adjusting and determining the length may include the following: i) having a set of connecting elements 190 of different lengths for the user to choose from; ii) having one connecting element 190 with a ratchet system such as a buckle, Velcro (registered trademark), or BOA (registered trademark) that can adjust the length; iii) adjusting the initial position of the anchor point relative to the standing (position), and even if the connecting element is kept at the same length, a similar effect to changing the length of the connecting element will be obtained. In the following figure, the BOA ratchet system reaches the shoulder strap at a position close to the chest so that when rotated, it can control the distance between both anchor points on the back of the wearable device 100. In this example, as the ratchet is rotated, the anchor point on the upper back can be shifted up and down to control the pre-tension of the passive element 190. The ratchet may also be fixed with a clutch to maintain the once-set state, or the clutch may be released so that it can move freely without applying force to the person.

[0248] b) Load of connecting elementLink elements may have specific loads (e.g., stiffness and damping) based on deformation. Figures 15 and 16 show how different stiffness values ​​of link elements provide different outputs. The procedure for adjusting the loads of link elements may include a set of link elements with different loads that the user or designer can choose from, or may include multiple elements that can be connected in parallel or series to modify the load (e.g., spring constant).

[0249] Human Subject Experiments: As a first step to demonstrate the functionality of this configuration, human subject experiments were performed with a passive wearable device 100 with an elastic linking element acting as a spring between both anchor members 110. Through preliminary human subject experiments on three healthy participants, we showed promising results. As shown in Figure 33, these studies showed that the proposed configuration has the potential to reduce muscle activity of key muscles of the back (erector spinae at L4 and T9 levels) and hip joints (biceps femoris, gluteus maximus) by up to 45% on average in three subjects while maintaining a constant posture at different angles while wearing the wearable device 100 compared to wearing nothing. Case in point #1B: Active back and lumbar support devices

[0250] While the device described in Representative Example #1A provides a supportive pulling force to the back joint that is passively generated by the user's body movements, the linking element 150 may be sized and designed to provide different types of support as described in Example 1a. However, to provide support tailored to different users or tasks, or to provide a higher level of support than a completely passive device, an actuator 120 may be used to control the position or force of the linking element 150.

[0251] Referring to FIG. 17, the device described in representative example #1B provides assistive pulling forces to the back and hip joints that are passively generated by the user's body movements, and the linking element 150 may be sized and designed to provide different types of assistance as described in example #1a. However, to provide customized assistance for different users and tasks or a higher level of assistance than a completely passive device, the actuator 120 may be used to control the position or force of the linking element 150. FIG. 18 presents a device that includes an actuator 120 component that can be used to control a cable attached between both anchor members 110. The actuator 120 may be used to control the position of the cable that generates the pulling force, or it may be used to directly control the force that the cable exerts on the user.

[0252] In this example, the system is the same as described in Example #1A, but in this case, instead of having an elastic element connecting both anchor members 110, a cable-driven actuation system is used. In this system, the user wears the actuator 120 on some part of the body (in this case the waist), a Bowden cable connects the actuator 120 unit to one side of the joint, and an internal cable connects to the actuator 120 at one end and to the other side of the joint at the other end. In this way, when the internal cable is actuated, a pulling force will be generated between both support members 110. In another embodiment, the actuator 120 may be directly connected to one side of the joint (without the need for a cover) and drive a cable connecting the actuator 120 at one end and the other side of the joint at the other end.

[0253] FIG. 18 shows an exemplary configuration for generating assistive torques around the hip and back joints. The device must be able to provide assistance to the hip and back joints during tasks such as lifting, maintaining a constant posture, transporting heavy equipment, or pulling / pushing, which are common activities that may result in injury over time, while being sufficiently transparent (like normal clothing) to other movements. An exemplary device may control impedance to make the device follow a prescribed trajectory or to realize a virtual load (stiffness, damping) that is a function of the joint angle. In this case, since the device spans both the hip and back joints, the impedance may be defined by the relative angle between the torso and the thigh. This angle may be measured by an on-board sensor 230 such as an IMU or a tension sensing element. The exemplary virtual impedance realized by the actuator is defined by the stiffness values ​​shown in FIG. 19, where F is the resultant force, theta is the relative angle between the thigh and the torso, and K is the realized stiffness. In this case, the supportive profile will feel like having a spring with stiffness K between the support members 110. The advantage of such an actuator-based profile over passive devices is that there is more flexibility in automatically configuring the support profile by selecting different impedances optimized for different movements and different users. Furthermore, the virtual impedance has the advantage that it can have different characteristics (realized stiffness, damping, etc.) for each phase of the movement, e.g., different impedance depending on the direction of the movement. Although moving down to grab an object may require less energy than moving up with the object against gravity, the change in resistance realized as the movement changes direction may provide more energy to the person moving up than moving down, which is not possible with passive springs.Additionally, by choosing a different "initial angle," the resulting profile can be shifted so that the force does not start until the user reaches that initial angle.

[0254] Figures 20A, 20B, and 20C show sample user motions where the joints move from an initial position of 0 degrees (standing) towards 90 degrees (bending the torso to a horizontal position) following a sinusoidal pattern. In particular, Figure 20A illustrates a representative stiffness profile that the device will achieve as a function of joint angle, Figure 20B illustrates a sample of the joint angles executed by a user following a sinusoidal motion as the torso is bent to lift a load from the ground, and Figure 20C illustrates the resulting force from that motion, which is the stiffness multiplied by the joint angle divided as a% of the lift cycle.

[0255] The example movement might be representative of bending the torso to grab an object from the ground. The resulting force is the product of the specified stiffness multiplied by the joint angle the user is executing. The force is shown for a% of the lift cycle, where 0% is defined as when the user initiates the movement, 50% is defined as when the user changes direction and the joint velocity is zero, and 100% is defined as when the user completes the movement by returning to a straight position.

[0256] In impedance controlled devices, the system may have different modes depending on the detected motion. Therefore, the on-board sensor 230 may be used to detect different activities and for each of these, a different predefined impedance profile may be selected. For example, there are multiple ways to lift a load, e.g. using a squatting technique or a crouching technique (legs straight), and the lifting may occur such that the object is in front of the body (back moves in a sagittal plane) or such that due to space constraints, the back must bend and twist during the lift. These different motions may be detected by an on-board motion sensor 230, such as an IMU, where an algorithm could define an initialization threshold (e.g., relative angle or angular velocity between the torso and femoral joints) that, when crossed, is considered as the initialization of the bending motion. The threshold may also be used to define whether the motion is occurring in a sagittal plane or if the user is bending and twisting the torso to lift. Different thresholds may be used to define whether the user is bending the leg during lifting. In another example, the on-board sensor 230 may be used for classification algorithms (e.g., neural networks, linear classifiers, nearest neighbors, decision trees, etc.) to detect which activity the movement is classified into. Impedances or assistive behaviors may be selected for each of these movements.

[0257] In another example, the device may be EMG controlled, where one or more EMG probes are placed on representative muscles that the device is intended to assist. Filtered EMG signals from the target muscles may be used to estimate the joint torques being produced by the target joints, or to use applied forces produced by the exoskeleton.

[0258] Motion initialization detectionIn some applications, the optimization algorithm may need to detect when an activity begins and ends so that it knows if the data collected during the iterations can be used to evaluate the target function. For example, for a device that supports the back during lifting, the optimization algorithm may need to split the lift into different phases: initializing the movement, moving to approach the object, grasping the object, and moving to an end position until the lift is completed. The optimization algorithm may evaluate an objective function for each of these phases. For example, the optimization algorithm may use on-board motion sensors 230, such as an inertial measurement unit, to detect changes in joint angles, velocity, or acceleration, to segment the movement and evaluate the objective function. The algorithm may detect changes in joint angles, velocity, acceleration, perform the movement initialization, and begin evaluating the objective function and measuring or estimating the objective indicators until the person completes the movement. Finally, when the acceleration or velocity changes, the objective function will be evaluated as the user moves to the final posture until the movement stops. This is recorded as one iteration of the movement, and the optimization algorithm evaluates the objective function to determine the next set of parameters to be tested as candidates for optimal support for the next movement iteration.

[0259] As one example, a control algorithm may consider different subgroups of activities and, for each of these activities, find a set of parameters that maximizes or minimizes an objective function.

[0260] The different main movements that the device supports may be divided into subgroups: for example, for a device supporting the back joints, these different subgroups may be: lifting an object in front with knee extension (bent lifting), lifting an object in front with knee bending (squat lifting), lifting an object to the side with a twisting movement of the trunk followed by a knee extension or squat technique (a common movement in several industries, such as construction, delivery, transportation, logistics, manufacturing, etc.), maintaining constant postures with different trunk angles (a common posture in surgery, caregivers, manufacturing, etc., which may result in back strain and ultimately injury).

[0261] The classification algorithm may use on-board sensors 230, such as motion, force, and pressure sensors 230, to classify movements into subgroups. For example, an IMU may be used to detect when a movement begins (change in joint acceleration or velocity) and may be used to detect the direction of the movement - for example, whether the torso is bending and / or twisting and the user is performing the movement in the sagittal plane (e.g., lifting an object in front of the user) or whether the torso is bending and twisting (e.g., lifting an object to the sides) and this may be used to classify the movements outlined above. Additionally, the synchronized movements of different body parts may be used to detect whether the user is crouching lifting (torso bent and legs straight) or squatting lifting (torso and legs bent).

[0262] The optimization algorithm will then evaluate whether the user is performing the movement and optimize the set of parameters.

[0263] As the user repeats activities that include movements in the subgroups described above, the parameters of the device may be fixed or may be automatically modified to maximize or minimize the objective function.

[0264] When a user performs an activity, the control algorithm will detect what type of movement the user is making and select the correct set of parameters based on the subgroup into which this activity is classified. Representative Example #1C: Robotic Garment Worn Under Clothing to Support Hip and Back Joints

[0265] As shown in Figure 35, the robotic garment is designed to be worn underneath the clothing and may consist of (i) a fully soft undergarment containing soft sensing components and force paths for comfortable transmission of assistive forces to the body, and (ii) a cable-driven actuation module that meets specific force, power, and bandwidth requirements and can be easily connected to the undergarment to power the distributed sensors and assist the user.

[0266] Wearable Device Structure For the robotic garment to be effective, it is important to design the appropriate force paths to transfer the tensile forces through the textile elements (which can be said to be the structure of the wearable device 100). The force paths across multiple joints can generate tensile forces to simultaneously assist these joints using a single actuation cable. This approach results in an optimized system that simplifies the requirements. FIG. 35 shows the structure of the wearable device 100 that can be used to offload both the back joints and the hip joints. The structure of the wearable device 100 is made of fabric parts that make up two joints, one at the upper back that is fixed to the body via the shoulder anchor members 110 (e.g., shoulder straps) and the other at the lower back via straps that are connected to the backs of both thighs. The connection elements 150 (e.g., actuation cables) directly or indirectly couple the shoulder and thigh anchor members 110. When actuated, the connection elements 150 actively generate tensile forces that simultaneously cross the user's back joints and hip joints. Because the force paths are offset from the biological joints (moment arms) of the thigh and back, these tensions generate assistive torques, unloading those joints.

[0267] Cable Driven Actuator 120An initial simplified quasi-static kinematics model was developed that shows the effect of the wearable device 100 structure with respect to key anatomical landmarks, including the estimated compatibility of the robotic garment-human interface, and actuation and transmission efficiency. One such module may be placed near the user's center of gravity (e.g., lower back or hips).

[0268] Sensors integrated into clothing : The robotic garment can incorporate sensors 230 to directly measure human motion, strap tension, interaction forces / pressures, and muscle forces on the fabric, without the need for IMUs, load cells, and EMG sensors on the fabric components. Fabric-enabled sensors 230 that can measure the above parameters can be integrated into the fabric to control the system to provide optimal assistance to the wearer and evaluate the biomechanical effects of using the robotic garment in an industrial environment. Sample soft sensors 230 include stretch sensors that measure motion tracking, soft tension sensors, in-insole pressure print sensors, or fabric-integrated EMG sensors to control and evaluate the robotic garment. Note that alternatively, the system can also include stiffness sensors such as IMUs, load cells, or EMG.

[0269] Integrating Sensors and Actuators into Robotic Clothing :For the development of the robotic garment concept, our team has in-house soft robotic prototyping processes that combine proven and novel garment manufacturing techniques. These processes allow us to create lightweight, modular, soft, conformal garments by assembling the aforementioned actuators 120 and sensors 230 as if they were seamlessly integrated into the fabric. As mentioned above, the robotic garments consist of a completely soft garment made of fabric with integrated load-transmitting force paths, sensors, and cable-driven actuation modules that can be worn underneath the garment and enclose the only rigid part of the system that can be easily attached to the undergarment.

[0270] This soft undergarment is worn under clothing and may contain some or all of the sensor components mentioned above: (1) stretch sensors 230 to measure joint motion, (2) tension sensors 230 to measure human-robot interaction forces, (3) EMG to measure the main muscles the device assists, and (4) pressure insoles in the legs that can be used to estimate GFR. These sensors 230 may be incorporated into the fabric components along with conductive fabric wiring that may route the various signals as if they were seams-free to a central connection location (where the actuation units are attached) located on the lower back.

[0271] The cable-driven actuation unit is placed near the user's center of gravity since this is the most advantageous location to place a heavy load. This module may integrate cable-driven actuation, PCB electronic components (high and low power electronics), and a battery. The user can connect the actuation unit to his back by sliding it into a semi-rigid slot panel located at the waist, and when the actuator 120 is placed, it makes a connection with the center panel to power the underwear part of the robotic garment and receive sensor information. The actuation cable is attached to the connection part of the underwear using a quick connect. The actuation module may also incorporate two IMUs in the rigid housing to add sensing capabilities, one in the actuation box and the other in the quick connect where the actuation cable is connected to the upper back. Furthermore, a vibration part may be incorporated into the mounting plate 220 so that haptic feedback can be integrated to alert the user, and this element may be used to inform the user that they are using a non-ergonomic posture when lifting or holding a static posture. Representative Example #1D: Optimized Control of an Active Back and Thigh Assist Device:

[0272] Activities that may require optimization include lifting objects in front with knees straight (bending), lifting objects in front while bending knees (squatting), lifting objects to both sides followed by torso twisting while bending or squatting (common in multiple industries such as construction, delivery, logistics, military logistics, manufacturing, etc.), and holding static postures at various torso angles (common postures for surgeons, caregivers, manufacturing, etc.). The proposed intuitive and adaptive controller 240 may optimize assistance for different users (bending and squatting with objects in front and to the sides, and holding static postures at various torso angles) with a focus on actions identified as important to reduce injury risk and improve productivity in different industries. FIG. 29A illustrates different activities that the device optimizes and assists. The optimization controller 240 may use human biomechanical metrics to optimize assistance parameters to minimize joint and basic muscle activity in these activities as much as possible without negatively affecting the wearer's natural posture during the targeted movement. The proposed controller 240 finds optimal assistance parameters for each considered movement for each user. For activities not included in the above categories (e.g., walking, running, stairs, etc.), the controller 240 may command the system to be completely transparent (no force applied to the person) so as not to restrict the movement.

[0273] The controller 240 includes one or a combination of the following: a load-controlled robotic garment, a classification algorithm that uses the on-board motion sensors 230 to distinguish between motions of interest, and an efficient optimization algorithm that optimizes device parameters (e.g., load, start) to maximize or minimize an objective function defined for each proposed task.

[0274] Load-controlled robotic apparel: We are developing a commercially available passive rigid exoskeleton and a soft passive device for unloading the back and hip joints. Our preliminary results indicate that passive loading is effective in unloading the target back muscles for some movements of some users. However, poor adaptation of the passive systems limits the applicability of these devices to different users and activities, and applying suboptimal assistance to unload the joints may negatively impact the wearer's ability and even hinder natural movements in some situations. Thus, in this example, we implement a controller 240 that optimizes the final load of the device and allows for initiation of movements for different movements and by different users. In this way, the device is programmed to provide virtual loads (e.g., springs, dampers) according to the biological joint movements. We leverage this experience to develop a load controller 240 such that the target force is a function of the desired virtual load (e.g., initiation, stiffness, damping) and the kinematic properties of the joints measured by the wearable sensor 230.

[0275] Classification Algorithms : An algorithm that uses wearable sensor information to classify the initiated movement to understand which activity the subject is performing, including lifting method (bending, squatting, front lifting, side lifting) and static posture, as shown in Figure 29A. The classification algorithm classifies the different types of activities the user is performing during the movement initialization phase, allowing to select different assistance parameters for each movement and optimize them independently. (See Figure 29B)

[0276] Optimization ControlAs explained above, the optimization algorithm can use measurements from various body-worn sensors 230 to efficiently optimize the control policy continuously online and improve important physical quantities such as minimizing the forces of the targeted muscles and minimizing deviations in the kinematic characteristics, dynamics and muscle activity of non-targeted muscles with or without the device. Our specific approach improves key metrics such as muscle activity of targeted joints and joint torque / power as the person performs a task without negatively affecting the wearer's natural biomechanics. Compared to existing approaches with predefined assistance profiles by manually adjusting parameters, or passive systems with a constant load for all people and all tasks, our approach allows a customized assistance profile to be automatically generated for the user after a short period of device use.

[0277] For the personalized wearable device 100 to be practical, the control algorithm must efficiently optimize the control parameters online while the subject performs a short, supervised task. In this example, we apply Bayesian optimization, a family of global stochastic optimization algorithms. Bayesian optimization algorithms are known to be highly sample-efficient. It also avoids many of the pitfalls associated with traditional gradient descent and parametric response surface methods (i.e., local minima, high bias).

[0278] Real-time data from the wearable sensors 230 can be used to estimate muscle activity of target muscles, joint angle positions, velocities, and interaction forces between the human and the robotic garment to optimize assistance (in terms of load and initiation).

[0279] After this period, initial control parameters are used and the device may begin to provide assistive forces as the user performs several repetitions of the target task. The algorithm then uses on-board sensing to continuously monitor EMG activity and the wearer's kinematic characteristics to optimize the degree of basic muscle unloading while ensuring that the natural kinematic and dynamic characteristics of the user and muscle activity are not undesirably affected. The parameters that define the properties of the virtual load and the onset of actuation are optimized for each activity considered, resulting in an independent set of parameters for each movement and each user.

[0280] Examples of objective functions that can be used in this task include:

[0281] Maximizing the degree of reduction (RMS and / or peak) of the main muscle groups of the back while minimizing the impact on the muscle activity of the abdominal muscles. The co-contraction of the abdominal and back muscles is crucial to ensure the rigidity of the structures that stabilize the back. Since the proposed device generates forces that mimic the action of muscles in the back, we hypothesize that the muscles in the abdominal area should maintain normal muscle activity compared to the baseline, so as not to negatively affect the risk of injury.

[0282] Maximizing the reduction (RMS and / or peak) of the main back muscle groups without negatively affecting posture: Maximizing the degree of deloading of the targeted muscles can be used alone as one goal or in combination with criteria to define posture quality. Posture quality can be defined in different ways, such as not changing the kinematic characteristics in a way that means the object is far from the body when lifting. The distance between the object and the body is considered one of the main factors that affect the risk of injury when performing lifting. Alternatively, a guideline that quantifies the risk of injury from an activity, such as the lift value, can be used as a metric. The lift value is an equation developed by NIOSH that evaluates how significant the risk of performing lifts (6), (100), and (101) is. This equation is a function of the horizontal and vertical distance of the body to the object, the asymmetric angle when lifting the object, and the frequency and duration of each activity. Any change in kinematic / kinetic characteristics compared to the baseline will result in a change in the lift value. On-board sensors can be used to evaluate the distance from the object to the body and the EMG activity of the back muscles. Changing your posture in a way that increases your risk of injury will ultimately force your muscles to work harder, and to do so, harnessing the muscular activity of the target muscles may be sufficient for your purpose.

[0283] Minimizing torque or power at biological joints. Peak and cumulative loads at joints increase the risk of injury (REF). Minimizing the amount of torque or power (RMS and / or peak) that biological joints must generate per repetition of a task is expected to weaken the biological muscles generating these torques. The following are alternatives to using EMG as a means of optimizing assistance that are worth investigating: Multi-parameter optimization, combining the goal of unloading joints as much as possible with posture-related metrics such as lift value or distance of object from body.

[0284] Additionally, robotic clothing could integrate vibration elements into the fabric that could be used to alert the user if they are in an unergonomic posture, which could be useful in a variety of situations, such as changes in movement characteristics due to fatigue or distraction.

[0285] The presented method optimizes parameters such as the onset and load of the assistance profile, which are a function of how the user behaves for each defined motion. Thus, even if the lift value is not identical to the conditions used for optimization, the final force based on the person's motion is somewhat adapted. Compared to existing approaches that manually adjust parameters and have a predefined assistance profile, or passive systems that have a constant load for every person and every task, this method allows a customized assistance profile to be automatically generated for the user after a short period of device use. This represents a significant improvement that could advance the field and understanding of how to develop an optimal controller 240 that adapts to individual physiological and neurological differences when performing various workplace activities. Representative Example #1E: Semi-Active Device for Back and Thigh Support

[0286] In other embodiments, the assist device may be configured to assist the back and hip joints as shown in Examples 1A and 1B, but may include a clutch system or non-incendive actuators 120 to control parameters such as:

[0287] a) Initial length of the connection element: The initial length (x01) of the passive connection element defines when tension begins to occur as the user bends with respect to a standing posture, and the distance between both anchor members when the user is standing is defined as x0. For example, if the initial length (x01) of the anchor element is greater than or equal to x0 (x01 >= x0), tension is generated as the user bends the torso or legs with respect to a standing posture for a certain activity. This connection element can be designed to be shorter than x0 (x01 < x0) if it is intended to provide tension while the user is standing. FIG. 16 shows how the output tension can be modified by defining different initial lengths of the connection element. The process of adjustment may include commanding the target position of the actuator 120 and using a sensor 230 such as an encoder or a potentiometer to define the current position of the actuator 120. Once the initial length is defined, the actuator 120 is controlled to stay in that position, eliminating the need to support the forces generated when assisting the joint or the need to be non-backdrivable, which results in energy savings compared to a backdrivable system holding the position while the assisting force is being generated.

[0288] b) Slack vs. Pretension Mode : If the device is intended to be completely transparent (applying no assisting force), the actuator 120 system or a clutch can be used to allow the system to slacken. In this case, the actuator 120 is commanded to release the cable so that there is sufficient slack (the operating cable is longer than the distance between the anchor members 110), disengage the clutch of the connection element so that no force can be generated, or open the pneumatic system so that the connection element does not apply force to the body.

[0289] c) Load on connecting elements : By using a clutch system or the actuator 120 to connect various elastic elements in parallel or to control the initial positions of various stretchable materials in parallel, a selection can be made between various loads to add or remove the load on the system. Representative Example #1F: Back Support Device with Semi-Rigid Elements

[0290] In another embodiment, the support device can be configured to support only the back joints, rather than supporting both the hips and back as shown in Examples 1A-1E.

[0291] In this case, the structure of the wearable device 100 consists of a waist belt that constitutes an attachment point directly on the lower back and a shoulder strap that creates an attachment point on the opposite side of the joint. The device may also include a semi-rigid part 200 extending between the first anchoring member 110 and the second anchoring member 110, the semi-rigid part 200 being flexible in bending but resistant to deformation under compressive forces. In another embodiment, as shown in the previous example, the semi-rigid part 200 can be used in combination with the anchoring member 110 that may be attached to the waist belt via a connecting strap, for example to provide additional support to the hip joint or to reduce the relative displacement of the waist belt with respect to the human body during actuation.

[0292] Semi-rigid component 200, in various embodiments, can include multiple flexion members 202 connected by multiple flexion links 204. For example, semi-rigid component 200 can include flexure hinges inspired by biological spinal construction methods (spinal discs and vertebrae) or rigid links (flexion members 202) connected by elastic elements (flexion links 204). In other embodiments, semi-rigid component 200 can be a semi-rigid structure 206, such as a foam, with the bending and deformation properties referenced above.

[0293] This connection structure can be completely passive by having elastic elements connecting to both ends of the structure, which generate tension in the structure as the user bends, and the semi-rigid structure 206 thus delivers the force at an angle to the user's back, as opposed to pure tension, which has the advantage that a smaller force magnitude is required from the user to produce the same force.

[0294] The connecting element 150 can be actuated such that a motor can control the relative position or the force between both anchor members 110. The motor can be used to set the level of pretension of the connecting element 150 (semi-active system).

[0295] 21 shows an example of a semi-rigid part 200 that includes two anchor members 110, one at the top and the other at the bottom. A connection element 150, such as an actuation cable or a resilient passive element 190, may be connected to provide an assist torque to the user's back.

[0296] One way to control active devices and adapt the assistance to the wearable device 100 to the individual and the task is to use optimization methods to optimize control parameters to maximize or minimize an objective function. Case Study #2: Knee Support Device

[0297] Some activities may require the user to remain in a squatting position for long periods of time. As an example, if a worker has to perform work in a confined space, they may need to remain in a squatting position to perform the work on their knees. Knee injuries may occur from maintaining an unergonomic position for long periods of time without sufficient recovery time. To mitigate this issue, the assistive device may provide torque to the knees so that the user's knees use less effort during that time. This may allow the worker to maintain this position without affecting their musculoskeletal health.

[0298] FIG. 22 shows an example of an embodiment of a soft wearable device 100 achieved by placing a cable driven actuator 120 on either side of the knee (front of the thigh and front of the calf) using a technique similar to that described in the previous example with anchor members 110. This actuator 120 can be attached to the shin or calf by wearing a calf anchor member 110 (e.g., a calf wrap device) around the shin and then attached to the top of the knee by attaching to a thigh anchor member 110 (e.g., a thigh wrap). When actuated, the cable provides a force that provides support for the knee during a squatting position. This device is useful to avoid knee injury when maintaining a squatting position for long periods of time. This activity involves a position that needs to be sustained over time.

[0299] Of course, the knee wearable device 100 can be configured in addition to or as an alternative to semi-active or passive support. For example, the knee wearable device 100 can include a semi-active actuator 120 (e.g., a clutch system, a non-back-driving actuator, or other suitable mechanism) configured to control how much tension is generated in the wearable device 100 by the wearer's movements or posture. To control the level of tension, the semi-active actuator 120 can be configured to control the length of the connection element 150 and / or to selectably engage one or more connection elements 150 of appropriate stiffness such that a desired tension is generated, as described above. Additionally or alternatively, the connection element 150 can include an energy storage device 190 (e.g., a spring), and the semi-active actuator 120 (e.g., a clutch) can be configured to selectively lock and unlock the energy storage device 190 to store and release energy absorbed by the connection element 150 from the wearer's movements or postures to assist the wearer. The passive wearable device 100 may include passive connection elements that directly or indirectly connect at least one torso anchor member 110 to at least one upper leg anchor member 110 and at least one lower leg anchor member 110 to provide knee support such that the wearer's movement or posture generates tension within the wearable device 100 and the wearable device 100 provides a moment (e.g., a restoring torque) about at least one knee joint of the wearer. Example #3: Shoulder Support Device for Overhead Work

[0300] In one example, a healthy subject may be required to perform overhead tasks to assemble various parts in a structure above their head. This is a very common task, for example in the automotive industry, where factory workers perform assembly tasks under cars. In one such case, the worker may perform multiple repetitions of grabbing various parts (such as screws, nuts, and other parts) from a table, stretching their arms above their head to attach the parts to the structure using power tools, and / or grabbing new parts from the table. Working in an overhead location puts strain on the neck and shoulder muscles. Shoulder muscles tend to fatigue very quickly when performing overhead tasks, which is why it is considered one of the most common causes of injuries resulting from body posture and reactions. 23, the wearable assistive device may be worn around the shoulder to provide assistance to the shoulder joint to prevent the shoulder joint from excessively heavy work when the user extends the arm over the head. In various embodiments, the wearable device 100 may include at least one anchor member 110 configured to be placed on the torso of a person wearing the wearable device 100, at least one anchor member 110 configured to be placed on the upper arm of a person wearing the wearable device 100, and at least one connection element 150 that directly or indirectly couples the at least one torso anchor member 110 and the at least one upper arm anchor member 110. In one embodiment, the active wearable device 100 for assisting with these tasks may include a cable-driven actuation system connected to the shoulder anchor member 110 and the arm anchor member 110 and extending between the shoulder and the arm. The actuator 120 may include an electromechanical motor, a Bowden sheath connecting the motor to one side of the shoulder joint, and an inner cable that connects one end to the motor and the other end to the arm. For example, as the motor moves, the connection element 150 (the inner cable) moves and helps move the arm up. The actuator 120 may have anchors under the arm attached to a garment piece that wraps around the arm and around the side of the chest, as shown in FIG. 23, so that as the cable moves, it generates tension that creates a supporting torque on the shoulder joint.

[0301] In some embodiments, the device may include one or more integrated sensors 230, such as an IMU on the arm to measure joint angles / velocities / accelerations, an IMU on the chest, one or more EMGs to measure muscle activity of the muscles the device is supporting, a pressure sensor in the actuator 120, and / or a pressure sensor in the joint between the user and the actuator 120 to measure the support force on the arm. These sensors 230 may be integrated into the wearable system.

[0302] A typical motion is shoulder movement from neutral to overhead position. A typical actuation profile may indicate initiation when the user reaches a particular shoulder angle, a predetermined stiffness of the actuator 120 units based on the arm angle (so that the actuator 120 provides assistance proportional to the arm angle, i.e., the greater the arm angle, the greater the support force), and a stop or offset of assistance when the arm starts to descend and the direction of rotational velocity changes.

[0303] Of course, the shoulder wearable device 100 can be configured in addition to or as an alternative to semi-active or passive support. For example, a semi-active shoulder support wearable device 100 can include a semi-active actuator 120 (e.g., a clutch system, a non-back-driving actuator, or other suitable mechanism) configured to control how much tension is generated in the wearable device 100 by the wearer's movements or posture. To control the level of tension, the semi-active actuator 120 can be configured to control the length of the connection element 150 and / or to selectably engage one or more connection elements 150 of appropriate stiffness such that a desired tension is generated, as described above. Additionally or alternatively, the connection element 150 can include an energy storage device 190 (e.g., a spring), and the semi-active actuator 120 (e.g., a clutch) can be configured to selectably lock and unlock the energy storage device 190 to store and release energy absorbed by the energy storage device 190 from the wearer's movements or posture to support the wearer. A passive wearable device 100 for shoulder support may include a passive connection element that directly or indirectly couples at least one torso anchor member 110 to at least one upper arm anchor member 110 such that a movement or posture of the wearer creates tension in the wearable device 100 and the wearable device 100 provides a moment (e.g., a restoring torque) about at least one knee joint of the wearer. Case Study #4: Modular Wearable Devices

[0304] The devices described in the previous examples may be combined to support multiple joints, such as knees, back, shoulders, or any combination thereof. As an example, as shown in FIG. 24C, a knee support device (FIG. 24A) may be combined with a back support garment and / or a shoulder support garment (FIG. 24B) to provide additional support for a given activity. As an example, due to workspace constraints, it may be necessary to perform overhead tasks while crouching in some environments, so a knee support device may be combined with a shoulder support device. These modules may be combined as if there were no seams, as the devices are purely soft and do not affect each other. Case Study #5: Clothing Integration

[0305] The proposed assistive device can be integrated into typical everyday work clothing. Figures 25A, 25B, 25C, 26A, and 26B show a sample embodiment of the integration of the wearable device 100 components. The figures show a system that includes work clothing with attachment points and a removable cable-driven sensor and actuator 120 device that can be clicked into the work clothing for periods of use.

[0306] Clothing integration is beneficial because it (i) provides the user with a platform familiar to everyday clothing, (ii) reduces user alignment errors since the clothing is difficult to put on incorrectly, (iii) eases hygiene concerns since fabrics that require machine washing due to sweat or foreign matter can be easily separated from non-washable sensors 230, electronics, and actuators 120, (iv) potentially obscures the fact that the user is wearing an assistive device, a pain point that persists from initial user research, and (v) reduces donning and doffing time of the system due to fewer parts.

[0307] FIG. 25A shows how the wearable device 100 components are used to wear the system over regular clothing. FIG. 25B and FIG. 26A show typical workwear with lightly padded sewn-in inserts that function as wearable device 100 components. Using a small, sealed Boa® system dial or Velcro® tab, the user can add pressure to these lightly padded inserts when they want to activate the system. When the system is not in use, the user can leave these inserts loose. FIG. 25C and FIG. 25B show the concept of using a hidden yoke or piece of fabric that is intended to provide support when the user tightens using the Boa® system dial or Velcro® tab. This design is particularly beneficial because it allows the wearable device 100 components to be hidden under the top layer of clothing during pressure application, which is an elegance and perception concern for some users. Example 6: Optimization of shoulder-assist control to support overhead tasks

[0308] In one example, a healthy subject may be required to perform an overhead task to assemble various parts within a structure above their head. This is a very common task, for example in the automotive industry, where factory workers perform assembly tasks underneath cars. In one such case, the worker may perform multiple repetitions of grabbing various parts (e.g., screws, nuts, and other parts) from a table, reaching their arms above their head to attach the parts to the structure using power tools, and / or grabbing new parts from the table.

[0309] Working at an overhead location puts strain on the neck and shoulder muscles, which tend to fatigue very quickly when working overhead, which is why they are considered one of the most common causes of injuries resulting from body position.

[0310] 27, a wearable assistive device may be worn around the shoulder to provide assistance to the shoulder joint when the user extends the arm over the head to avoid excessive strain on the shoulder joint. In some embodiments, the device may include one or more integrated sensors 230, such as an IMU on the arm to measure the angle / velocity / acceleration of the joint, an IMU on the chest, one or more EMGs to measure muscle activity of the muscles it is assisting, a pressure sensor in the actuator 120, and / or a pressure sensor in the joint between the user and the actuator 120 to measure the support force on the arm.

[0311] A typical motion is a shoulder movement from a natural position to an overhead position. A typical actuation profile may detect initiation when the user reaches a threshold shoulder angle, adjust a predefined stiffness of the actuator 120 unit based on the arm angle (such that the higher the arm angle, the greater the support force, so that the actuator 120 provides assistance proportional to the angle of the arm relative to the initial angle considered as the onset of the overhead motion), and deliver a stop assistance or offset when the arm starts to descend and the direction of rotational speed changes.

[0312] An optimization algorithm can be used to select an appropriate amount of assistance from the wearable device 100 to avoid overworking the muscles that the device is configured to target. For example, the objective function can be a peak for a target muscle, such as the upper trapezius, in which case the objective can be to minimize this value to avoid overworking that particular muscle while performing a task.

[0313] Suitable parameters for optimization include, but are not limited to, the offset and stiffness of the actuation profile as a function of joint angle. A stiffer actuator 120 may mean that the wearable assistive device feels more supportive, but at the same time, too much stiffness may not provide optimal results, since a spring that is too strong may mean that the user needs to provide resistance against this stiff element to avoid performing too fast or uncontrolled movements. The offset of the actuation according to the arm angle is also an important parameter, since the actuator 120 supporting the arm may need to become transparent or reduce support as the user begins to lower the arm. An example of a simplified assist profile based on stiffness is shown in FIG. 27. FIG. 27 shows an exemplary graph of the actuator 120 force versus the arm angle. The simplified actuation assist shows that the actuator 120 force, and therefore the support to the shoulder, is stiffness provided in proportion to the arm angle.

[0314] Different individuals may perform tasks in different ways. Moreover, while wearing an active device, the wearer may react in different ways to the assistance provided by the device and the optimization associated with the device. Using an algorithm that automatically optimizes the device's assistance (e.g., stiffness) parameters based on the optimization of the objective (e.g., muscle activity), different individuals may receive improved benefits from the system, and the device may be automatically customized for slightly different activities. In this way, an active wearable shoulder (or upper limb) device 100 that uses an online optimization process to optimize a parameterized assistance profile and maximize an objective function can be used for different activities and different users. The wearable assistive device can automatically adjust the assistance parameters according to the optimization techniques described herein.

[0315] The optimization process may evaluate objectives based on initial settings or search points, update the profile using collected data, determine next candidate search points, evaluate objectives with these settings, check convergence criteria, etc. An example of this optimization is shown in Figure 28.

[0316] As described above, methods similar to the two-parameter Bayesian optimization example can be used to find optimal assistance parameters for various objectives. For example, in one embodiment, a Bayesian approach can be used to optimize one or more operating parameters when the relationship between the operating parameters and the objective function is unknown or complex.

[0317] The following pseudocode shows this example process using Bayesian optimization. [Table 1-1] [Table 1-2] Example 7A: Optimization of back support control to support lifting of loads

[0318] With reference to Figures 29A, 29B, and 30, we provide further examples of small to moderate levels of assistance (20-30% of typical joint torques during comfortable walking), combined with the low inertia and non-restrictive nature of the proposed robotic clothing platform and existing new opportunities for wearable robot control where assistance needs to be synchronized with the wearer's underlying muscle function. We hypothesize that by providing appropriate timed assistance to the back and hip joints in the range of 20%-30% during tasks such as lifting or holding a static posture, industrial workers can improve their performance and reduce the risk of injury.

[0319] The control algorithm may optimize device parameters based on EMG measurements and joint kinematics / dynamics. Real-time data from the sensors 230 may be used to estimate muscle activity of target muscles during activity, joint angle positions, velocities, and interaction forces between the human and the robotic garment, and this information may be used to optimize the assistance profile (in terms of loading and onset of actuation). During a monitored period of device use, sensed information on the wearer's neutral kinematics and muscle activity is collected as a baseline (when no assistance is being applied). After this initial period, the user wears the device with an initial pre-determined load based on the relative angle between the torso and thighs. The algorithm then continuously monitors the EMG activity and the wearer's kinematics to optimize the degree of basic muscle offloading while ensuring that the user's natural kinematics, dynamics, and muscle activity are not undesirably affected.

[0320] The following actions are cited as being critical to optimize: lifting an object in front while keeping knees straight (bending), lifting an object in front while bending knees (squatting), bending and squatting followed by torso twisting movements to both sides (common actions in multiple industries including construction, delivery, logistics, military logistics, and manufacturing), and holding static positions with various torso angles (common positions for surgeons, caregivers, manufacturing, etc., which can lead to back strain and ultimately disability) The algorithm in this case consists of the following elements:

[0321] Classification algorithms to distinguish the above mentioned movements: on-board sensors 230 (inertial measurement unit) to classify the different movements into subgroups. This classification is developed using detection of changes in joint angles, velocities or accelerations to define when the movement started, detection of the direction of the movement (trunk bending in the sagittal plane (frontal lifting) or trunk bending and twisting (side lifting)) to define whether it is a bend or squat (legs moving in sync with the trunk or legs not moving (segmental velocities below threshold)).

[0322] An optimization algorithm that optimizes device parameters to maximize or minimize a defined objective function.

[0323] The subject performs multiple repetitions of the described movements while the device is not in motion to collect baseline kinematics, dynamics, and muscle activity. The subject then performs multiple repetitions of each movement while an optimization algorithm uses on-board sensors to optimize device parameters to maximize or minimize a desired objective function. Passive loading has proven effective in unloading the dorsal joint under predefined movements for some users, but the inability to adapt passive devices limits the applicability of these devices to different users and movements. In this case, the properties of the device load (stiffness, damping, etc.) and the onset of actuation are optimized for different users and movements. The device is load-controlled so that it can act as a virtual load (e.g., spring, damper) according to the movement of the biological joint. The parameters that define the properties of this virtual load and the onset of actuation are optimized for each activity considered, creating an independent set of parameters for each movement and each user.

[0324] Examples of objective functions that can be optimized include:

[0325] To maximize the degree of reduction (RMS and / or peak) of the main muscle groups of the back without affecting the activity of the muscles in the abdominal region. As the proposed device generates forces that mimic the behavior of muscles in the back, co-contraction of the abdominal and back muscles is important to ensure the rigidity of the structure, and maintaining normal activity of the abdominal muscles is considered important to not negatively affect the risk of injury.

[0326] Maximize the reduction (RMS and / or peak) of the main muscle groups of the back without affecting the kinematic characteristics in a way that implies that the object is farther away from the body when lifting. On-board sensors are used to assess the distance from the object to the body and the EMG activity of the back muscles. The distance between the object and the body is considered one of the main factors that influence the risk of injury and implies an increased force to hold or lift the load.

[0327] Maximize the degree of back muscle unloading (RMS or peak) without increasing lift value - Lift value is an equation developed by NIOSH that evaluates how significant the risk of performing a lift is. This equation is a function of the horizontal and vertical distance of the body to the object, the asymmetric angle of the object when lifting, and the frequency and duration of each activity. The resulting index is used as part of the goal to minimize the risk of injury, and any change in the movement characteristics will result in a change in the lift index.

[0328] Our goal is to make these systems usable in real industrial environments where tasks are not structured. We assume that the proposed method optimizes parameters such as the onset and load of the force profile, which are functions of how the user moves for each predefined action that the device is intended to assist. After the optimization finds the optimal load and onset for each proposed activity, we expect some adaptation of the final forces based on the human movements, even if the lift values ​​are not identical to the conditions used for the optimization. Example 7B: Optimization of back support control to support lifting of loads

[0329] In some environments, workers are required to bend over when performing tasks. This puts strain on the lower back, which can lead to chronic back pain and more serious long-term back disorders. A device similar to that described in Example 1 can be attached to the body to provide support for the lower back. In this case, the assist device consists of an actuator 120 attached to the body via two anchors 110, one at the lower back and a second around the thorax. A sample system may include an inflatable device behind the back to provide support to the back when the user is lifting objects, bending forward, or reaching for objects. Such a device may reduce psoas muscle activity.

[0330] In another embodiment, a cable-based actuation approach for the back can be used, such as that described in Figure 21. One or more anchor members 110 can be used for the torso and lower back, each of which can have a cable attached. When actuators 120 apply a force, a torque can be generated to reduce the effort required by the back muscles when performing several different tasks described herein, such as lifting and carrying.

[0331] As an example, this application focuses on a repetitive task where a user needs to pick up objects from the ground and place them on a shelf in front of them, in an environment where the operator must bend forward to pick up these objects.

[0332] An example device may include a combination of sensors 230 such as a chest IMU to measure torso angle, one or more EMGs to measure muscle activity of target muscles, and a pressure sensor to measure the degree of support of a back support device.

[0333] An optimization algorithm may be set to optimize the objective function. Exemplary suitable objectives may include peak EMG of the lumbar or thoracic erector spinae muscles, user self-assessed comfort using the handle or input device, time to perform each lift, and posture. For example, a device that provides knee support during a squatting position may improve the posture of the wearer depending on how much support the device provides. A perfect posture may be defined in one embodiment as a posture in which the user keeps their back as straight as possible while performing a task, or as close to straight as possible when not wearing the device (back reference angle).

[0334] In this case, the objective function can include one or more of these parameters using a weighting function. For example, the objective function can include objectives related to the user's self-assessment and the time of the task mentioned above, with each parameter having a weight of 0.5 so that both subjective (user-dependent) and objective measures (prediction of task capacity) are included. In this application example, the user's choice of comfort can be rated as a value between 0 and 1, and the time to perform each lift is normalized to a number considered as the maximum lift, which is also between 0 and 1. Since both objectives have different units, normalization is applied as described so that both measures can be added using a weighting function.

[0335] In this application, the user can be prompted with a handle to select a comfort level when performing a task. The time to perform a task can be defined as the time elapsed between when the user starts bending their torso to reach for an object and when the user returns to the starting position. The back IMU can be used to detect changes in the velocity or acceleration of the back to detect when the user is starting the motion and, once the lift is complete, the user decelerates their torso back to a more straight position similar to where they started. This can be estimated using the back IMU by looking for changes in velocity, acceleration, and angle.

[0336] The actuators 120 can be controlled to provide different amounts of assistance during task execution. In this application, the system is providing lower back support during a task involving trunk bending, so assistance can be delivered as a function of trunk angle measured by a sensor such as an IMU. To do this, the assistance profile can be parameterized as a second order polynomial f=kθ2. Figure 31 shows an example of a representative assistance profile by varying the parameter k. A higher stiffness value means the back receives more assistance as a function of trunk angle during the bending motion.

[0337] In one embodiment of an optimization algorithm to assist this movement, the algorithm may have the objective of optimizing body posture while controlling the amount of assistance. Since it is beneficial to keep the back as straight as possible while lifting a load according to ergonomic guidelines, the optimization algorithm may use back angle as the optimization objective. In this application, the algorithm may adjust the stiffness of the actuator 120 while optimizing the back angle relative to a straight posture. In this application, the variable being optimized is the amount of stiffness in the actuator 120, and the objective is the peak torso angle while performing the task. In situations where a worker is performing a task for the first time, it is particularly useful to assist as much as possible while at the same time promoting proper body posture while performing the task. Example 8: Optimization of back support control to support prolonged standing

[0338] Holding a static posture for a long period of time can put some areas of the body under tension. Back disorders or fatigue can occur by maintaining a posture for a long period of time without sufficient recovery time. To alleviate this issue, an assistive device can provide torque to the back so that the user's back exerts less effort during that time. This allows the worker to maintain this posture without affecting their musculoskeletal health.

[0339] In this application, a device such as that shown in Fig. 21 can be useful as described above. If the actuator 120 provides a force, it can provide back support during standing. This device can be useful to avoid back disorders when maintaining a standing position for long periods of time.

[0340] Suitable optimizations may include device power consumption, EMG of muscles around the back joints, and overall posture. For example, a device that can provide back support when the wearer is in a standing position can improve the wearer's posture depending on how much support the device provides. In this case, perfect posture can be defined as the user standing upright (back as straight as possible) while performing a task, or as close to straight as possible when not wearing the device (back reference angle).

[0341] The objective function is to provide as much assistance as possible without affecting the wearer's overall posture compared to the posture the wearer normally has when performing the same task. The device can integrate sensors 230, such as IMUs on the thighs and torso, that measure the back angle. When the device is set in a fully transparent mode (i.e., not activated) but turned on, the system can perform the task in the same way as when the user is not wearing the device, but at the same time, it can measure and record sensor and device information. In this fully transparent situation, the device records the back angle to use as a reference when assistance begins. In the next step, the actuators 120 apply an initial tension (medium tension) and the optimization algorithm can adjust the level of assistance while minimizing the error between the joint angles relative to the angles the user had when the device was in full transparent mode.

[0342] This optimization is useful in ensuring that the device is providing as much assistance as possible without changing the overall ergonomics of the task, given the user's experience performing this activity. In other situations where it is not ideal to start with ergonomics, the device can be configured to minimize the error between the user's joint angles when performing the task and a reference, such as the OSHA recommended body posture when performing that task. This same concept can also be applied to lifting and carrying loads, etc. Example 9: Optimization of back support control to support carrying a load

[0343] As described in Example 3 above, this method can also be applied to carrying a load, where again appropriate optimizations may include device power consumption, EMG of muscles around the back joint, and overall posture. For example, a device that can provide back support when the wearer is carrying a load can improve the wearer's posture depending on how much support the device provides. In this case, perfect posture can be defined as the user standing upright (back as straight as possible) while performing the task, or as close to straight as possible when not wearing the device (back reference angle).

[0344] The objective function is to provide as much assistance as possible without affecting the wearer's overall posture compared to the posture he or she normally has when performing the same task. The device can integrate sensors 230, such as thigh and torso IMUs, that measure the back angle. In this case, the objective can be to minimize the error between the joint angles and the OSHA recommended body posture when performing the task.

[0345] This optimization can be useful to ensure that the device is providing as much assistance as possible by providing maximum assistance to ensure overall ergonomics. Example 10: Optimization of hand assist control to support grasping of a tool

[0346] Gripping parts for hours a day can be a hazard, and the added force of the hands increases the risk of hand and wrist injuries. Gripping for extended periods of time is a frequent cause of hand and wrist injuries for employees. Tools need to be held firmly enough to maintain control, but not too tightly. Gloves that reduce static gripping force can help avoid injury and excessive strain. Static gloves with high friction materials have been used to improve grip on power tools, but active devices that provide gripping assistance when needed and reduce the amount of force a user must exert to maintain control and a firm grip on the power tool may enhance work performance and mitigate the risk of injury.

[0347] Several tasks may require additional hand or wrist support, and several illustrative examples are provided below.

[0348] 1. For example, to hold something for a long period of time.

[0349] 2. Repeatedly pressing a button or pressing a trigger to activate a tool. For example, if you have a drill, the user needs to press a trigger to activate it. Pressing a tool trigger once requires very little effort from the user, but if a tool trigger needs to be pressed multiple times, it can lead to fatigue and other complications such as carpal tunnel syndrome.

[0350] 3. Handling objects such as tightening equipment by hand.

[0351] As shown in FIG. 32, an example of an assist device can include one actuation unit on the back of each finger that can provide assistance to each finger when compressed. A common activity in manufacturing that would benefit from such a device is holding a power tool for extended periods of time. The hand assist device can include an anchor member 110 located on the distal portion of the finger (e.g., the tip of the finger as shown in the figure) and another anchor member 110 located on the wrist (as shown) or other suitable body part (e.g., the proximal portion of the finger; base of the finger or palm) located proximal to the joint to be supported (e.g., the knuckle of the finger).

[0352] Optimization algorithms can be applied to assist healthy users' hands when performing tasks such as holding a power tool for long periods of time or pressing the tool's trigger multiple times per day, which are repetitive or sustained activities that can result in excessive strain and other effects.

[0353] Typically, the index finger is used to press a trigger in the tool, and the remaining fingers are used to grasp and hold the tool.

[0354] This application focuses on long-term sustained gripping of a tool.

[0355] An example of an objective for this optimization problem is related to wearer forces such as muscle activity, in this case the peak muscle activity of the finger flexors during gripping.

[0356] The support glove may include one actuator 120 per finger, which is integrated into the glove that is secured on one side to the tip of each finger and on the other side to the wrist.

[0357] The appropriate parameter to control for this optimization problem is the magnitude of actuation: the device provides more actuation to reduce peak EMG activity.

[0358] A suitable objective function for this problem is the grasp stability, or grasp quality. There are many functions that can be good proxies for improving grasp quality, such as the sum of the components of the force applied perpendicular to the object boundary, which indicates the internal force the object will withstand when an external obstacle is applied. Thus, the quality measure can be defined as the modular sum of the perpendicular components of the forces applied to the object. The optimization algorithm can use this metric as an optimization variable. The device includes a sensor 230 that measures the perpendicular force applied to the device beneath the surface of the finger.

[0359] However, since a worker may hold the power tool for an extended period of time, a second objective is battery usage, defined as estimating the level of battery remaining if the worker continues to hold the tool for a preset amount of time (e.g., 2 hours). To measure this, the amount of power the device is currently using and the power remaining in the battery can be measured by taking a battery voltage reading. The battery objective is to maximize the battery remaining after 2 hours of use.

[0360] An objective function might give relative weights of 0.7 and 0.3 to grip quality and predicted battery life after 2 hours of use, respectively, except that when the predicted battery life falls below 10% after 2 hours of use, the weights of each objective change to 0.4 and 0.6.

[0361] The actuation interval can be parameterized to vary the amount of pressure applied to each finger to optimize the above objectives.

[0362] To solve this problem, optimization algorithms such as Bayesian optimization and gradient descent can be applied. Example 11: Optimization of knee assist control to support squatting posture

[0363] Some activities may require the user to remain in a squatting position for long periods of time. As an example, if a worker has to perform work in a confined space, they may need to remain in a squatting position performing kneeling tasks for long periods of time. Knee injuries may occur from maintaining an unergonomic posture for long periods of time without sufficient recovery time. To alleviate this issue, an assistive device may provide torque to the knees so that the user's knees use less effort during that time. This may allow the worker to maintain this posture without affecting their musculoskeletal health.

[0364] Referring back to FIG. 22, an example of an embodiment of a soft wearable device 100 is shown, which is achieved by placing a cable driven actuator 120 on either side of the knee (front of the thigh and front of the calf) with a technique similar to that described in the previous example with anchor members 110. This actuator 120 can be attached to the calf by wearing a calf wrap device around the shin and then to the top of the knee by attaching it to the thigh wrap. When tensioned, the actuator 120 provides a force that provides knee support during a squatting position, which is useful when maintaining a squatting position for long periods of time to avoid knee injury. This activity involves a position that needs to be maintained over time, which is useful for an optimization algorithm to optimize the amount of support the device provides to the knee.

[0365] Suitable optimizations may include device power consumption, EMG of muscles around the knee joint, and overall posture. For example, a device that can assist the knee when the wearer is in a squatting position can improve the wearer's posture depending on how much assistance the device provides. In this case, perfect posture can be defined as the user standing upright (back as straight as possible, knees as straight as possible) while performing the task, or as close to straight as possible when not wearing the device (reference angle of back and knees).

[0366] The objective function is to provide as much assistance as possible without affecting the wearer's overall posture compared to the posture the wearer normally has when performing the same task. The device can integrate sensors 230, such as IMUs, on the calves, thighs, and back to measure the angles of the knees and back. When the device is set in a fully transparent mode (i.e., not activated), but turned on, the system can perform the task in the same way as when the user is not wearing the device, but at the same time, can measure and record sensor and / or device information. In this fully transparent situation, the device records the angles of the knees and back to use as a reference when assistance begins. In the next step, the actuators 120 are tensioned with an initial tension (medium tension) and the optimization algorithm can adjust the level of assistance while minimizing the error between the joint angles relative to the angles the user had when the device was in full transparent mode.

[0367] This optimization is useful in ensuring that the device is providing as much assistance as possible without changing the overall ergonomics of the task, given the user's experience performing this activity. In other situations where starting with ergonomics is not ideal, the device can be configured to minimize the error between the user's joint angles when performing the task and a reference, such as the OSHA recommended body posture when performing that task. This same concept can be applied to lifting, carrying loads, etc. Example 12: Optimization of elbow assist control to support lifting a load

[0368] When lifting an object, elbow support may be required to hold the weight of the object. A device similar to that shown in FIG. 27 can be worn at the elbow by attaching one end to the forearm and the other end to the arm. When tensioned, the device can provide assistance to the elbow joint, thereby minimizing the level of effort the user needs to exert.

[0369] Suitable optimizations may include power consumption of the device and EMG of the muscles around the elbow joint, for example a device that can assist the wearer's elbow when lifting a load could use the peak EMG activity of the muscles around the elbow joint as the objective function.

[0370] The device may incorporate sensors 230, such as IMUs, on the arm and forearm to measure elbow angle, and may also include EMG to measure muscle activity of target muscles in the elbow. A suitable parameter to optimize may be the level of assistance.

[0371] An example parameterized support can be defined as a load function (stiffness) based on elbow angle. The load factor can be optimized by an algorithm to maximize or minimize an objective function.

[0372] This optimization can be useful in ensuring that the device is providing as much assistance as possible while reducing the EMG muscles of interest. Wearable devices to support the spine and / or hip joints

[0373] Often people perform tasks that can lead to fatigue, reduced productivity, and ultimately increased disability. Injuries and fatigue from excessive heavy work pose economic and public health problems in industrial (warehouse, logistics, manufacturing, care, etc.), sports, consumer, elderly, and other applications. Back pain, fatigue, and injury are common disabilities that have a significant impact on our lives and our ability to effectively work. A wearable device 100 that provides support to the wearer's joints and assists with tasks and postures such as lifting, reaching, and static postures, thereby reducing fatigue, improving productivity, and / or reducing the risk of injury, can have a significant impact on productivity, enhancing the ability to do more, reducing fatigue, and mitigating the risk of injury.

[0374] FIG. 36 illustrates the concept of a device that may be worn to assist a user's motion or posture during lifting and / or reaching tasks. These tasks are common in industrial environments (e.g., warehouse, logistics, automotive, construction, care), elderly, consumer, recreational, and sports. The robotic garment can provide assistive force when needed (e.g., lifting and reaching tasks), but also be completely transparent when desired (e.g., while driving a truck, eating lunch, during breaks) and be unnoticeable during tasks that do not require assistance. In other words, the robotic garment can assist when needed, but be completely unnoticeable otherwise, potentially wearable throughout an entire workday.

[0375] The robotic garment can be configured to assist one or more joints by fastening to either side of the target joint, providing the wearer with assistive tension that provides support during movement and / or posture. In one embodiment, the active device can assist both the back and hip joints by providing tension across both joints, inspired by the way muscles and tendons generate force during tasks such as lifting.

[0376] FIG. 37A shows how tension can generate assist torque across the back and hip joints. The tension traverses the joints at a distance from the center of rotation (moment arm), which generates torque around the joint to assist joint movement and / or posture and reduce musculoskeletal fatigue and stress. In one example embodiment, tension across the hip and back joints can be translated into assist torque to assist hip and back movement and / or posture. This system configuration is useful for assisting the wearer during numerous activities such as lifting, static postures, reaching, etc.

[0377] FIG. 37B shows an example of a fully portable device capable of supporting the back and hip joints according to the structure shown in FIG. 37A. The device may include fabric pieces that wrap around the user's thighs and shoulders, and one fabric piece that wraps around the waist to allow the system components to be secured to the body. In the current implementation, the system may include two anchor members 110 (one for each leg), both of which may be connected via a load-balancing strap 112 to create one anchor member 110 (Anchor 1) at the lower back. The load-balancing strap 212 may slide relative to the anchor members 110 via elements such as low-friction rollers, bearings, pulleys, low-friction buckles, etc., as shown in FIG. 13. The load-balancing element 212 allows the user to move the legs freely during tasks such as walking, and creates tension in the torso and / or wearable device 100 during tasks that involve bending both legs (lifting, reaching, etc.). Additionally, this load balancing element 212 distributes tension to both legs by splitting the load to both anchor members 110. At the top of the torso, shoulder straps wrap around the wearer's shoulders and create anchor points at the upper back as shown in FIG. 37B. Additionally, conformal fabric elements (e.g., padding, foam) may be added to the back system to maximize comfort and moment arms (the relative distance between the joint's center of rotation and the tension of the wearable device 100). This element may include pockets to organize system components such as batteries and electronics.

[0378] An actuator 120 component, such as a cable drive system, may be connected between two anchor members 110 (Anchor 1 and Anchor 2) in the system. When the motor rotates, the connection element 150 wraps around a pulley, generating a force between the two anchor members, as shown in Figures 34A, 34B, 34C, and 38. It should be noted that alternative approaches and anchor members 110 should be considered, for example, where two motors can be used to provide tension to the back joint and hip joint separately. Considering the synergistic effects between the hip joint and back joint during lifting and static posture, we will focus here on describing an embodiment that supports both joints simultaneously. This has the advantage of reducing weight, complexity, and cost compared to solutions that provide support to both the lumbar and back individually, i.e., independently support the lumbar and back. It should also be noted that alternative actuation techniques may be used to generate tension across the different anchor members 110, such as fluids (pneumatic, hydraulic), electroactive materials, electromechanical actuation systems, cable driven actuation systems, alternative transmission materials (e.g., string, cable, ribbon, tape), clutches, etc.

[0379] FIG. 38 shows the detailed design and implementation of the electromechanical actuator 120 system, including the motor 121 and the cable transmission system (a series of guides 127). As an example, the system incorporates a T-motor, a motor commonly used in consumer applications (e.g., drone industry), which may provide higher capacity and lower cost than other possible options. The shaft 122 (not shown) of the motor 121 is connected to a first rotating pulley element 123 (input belt pulley) that connects to a second rotating pulley element 124 (output belt pulley) using a timing belt 126, the ratio of the sizes of both pulleys 123 and 124 being adapted to the initial gear reduction of the system (2:1 in this case). When the motor 121 rotates, this motion is transferred to the output pulley 124 via the timing belt 126. The connecting element 150 wraps around a third rotating pulley element (e.g., a spindle) that is connected to the output belt pulley 124.

[0380] Further reduction in speed is achieved through a series of guides 127 (e.g., pulleys or cable pulleys) as shown in FIG. 38. As previously mentioned, this concept serves as an alternative to a gearbox that captures the benefits of gear reduction without some of the drawbacks such as friction, inefficiency, and noise. FIG. 38 shows how passing the actuating element 150 around rotating or low friction elements achieves the equivalent reduction in speed to having a gearbox with a (n+1):1 ratio (where n is the number of rotating or low friction elements). In this case, five pulleys are used to achieve a reduction of 6:1. The advantage of incorporating a 6:1 ratio in this manner is that by adding this 6:1 reduction, the pulleys can be made smaller in size, which aids in the implementation and size of the actuation.

[0381] 38, the actuator 120 may include a first planar member 128 coupled to the first anchor member 110 and a second planar member 129 coupled to the second anchor member 110. A plurality of guides 127 (e.g., rotating pulley elements, low friction grommets or studs, etc.) may be disposed on the first planar member 128 and the second planar member 129, and a connecting element 150 (e.g., a cable, ribbon, etc.) may alternately pass through opposing guides 127 on the first planar member 128 and the second planar member 129 to couple the first planar member 128 and the second planar member 129. A first end 151 of the connecting element 150 coupling the first planar member 128 and the second planar member 129 may be fixed to the first anchor member 110, and a second end 152 of the flexible elongated element 150 may be coupled to the motor 121. The actuation motor 121 pulls on the second end 152 of the connecting element 150, thereby pulling the first anchor member 110 and the second anchor member 110 towards each other.

[0382] In another embodiment (not shown), the guide 127 may additionally or alternatively be directly coupled to the first anchor member 110 and the second anchor member 110 rather than to the planar member. One skilled in the art will appreciate that the planar member structure allows the actuator 120 to be selectively attached and detached from the wearable device 100.

[0383] Referring to Figs. 39A, 39B, 40A, 40B, 41A and 41B in front, an alternative cable driven actuation system design is shown. As shown, this actuation element includes the quick attachment feature of the wearable device 100, and the design allows for fabric parts to be attached to a machined mechanism. As shown in Figs. 39A and 39B, this version can be fully enclosed with a cover. Figs. 40A and 40B show the detailed mechanical design and various components that make up the system. The electromechanical actuator 120 system can include a brushless DC motor and a cable-based or ribbon-based gearbox system. As an example, the system incorporates an outrunner motor, such as T Motor's U7 280kv motor, which is commonly used in consumer applications (e.g., drone industry), and can provide higher capacity and lower cost than other possible options. The motor shaft is connected by a timing belt to an input belt pulley which connects to an output belt pulley, the ratio of the sizes of both pulleys being adapted to the initial gear reduction of the system (7.75:1 in this case). The actuation cable is wound around a cable spindle which is connected to the "output belt pulley". Further reduction can be achieved through one pulley (the "cable pulley"). As mentioned before, this concept serves as an alternative to a gearbox that can capture the benefits of gear reduction without some of the drawbacks such as friction, inefficiency, weight, cost and noise. Figures 41A and 41B show a practical system according to the design described in Figures 39A, 39B, 40A and 40B. Referring back to Figure 34B, by passing the actuation element around a rotating or low friction element, a reduction equivalent to having a gearbox with a ratio of n:1 (where n is the number of parallel cable sections) can be achieved. In this case, two parallel cables are used to achieve a further reduction of 2:1. The advantage of incorporating a 2:1 ratio in this manner is that by adding this reduction it is possible to keep the size of internal reductions such as timing belt transmissions small, which benefits actuation packaging, mass, cost and size.

[0384] 40B, the actuator 120 may be coupled to the first anchor member 110 and may include a motor 121 and a first rotating pulley element 123 connected to an output 122 of the motor 121. A second rotating pulley element 124 may be coplanar with and radially offset from the first rotating pulley element 123 and coupled to the first rotating pulley element 123 via a timing belt 126. A third rotating pulley element 125 may be coaxially disposed with and axially offset from the first rotating pulley element 123 and fixedly coupled to the first rotating pulley element 123. A connecting element 150 may be wrapped around the third rotating pulley 125 and a second end 152 or a middle portion 153 of the connecting element 150 may be coupled to the second anchor member 110. The actuation motor 121 shortens the connecting element 150, thereby pulling the first anchor member 110 and the second anchor member 110 towards each other.

[0385] FIG. 42 shows a device including a cable-driven electromechanical actuation system worn by a user. This system is described above and integrates the actuation system shown in FIG. 38. In this embodiment, the actuator 120 is located on the upper torso and is fixed to the upper torso where both shoulder straps connect, the end of the cable 150 is connected to the load balance strap 112, and when the motor rotates, tension is generated between the shoulder straps and the load balance element 212 as described above (FIG. 37A), which can generate torque across the hip and back joints as needed. Another advantage of this system is that when the cable goes "slack", no force is transmitted to the user. As shown in FIG. 42, the above-mentioned system can be worn over clothing to provide support to the back and hip joints during lifting and holding a static position. FIG. 42 shows details of an embodiment in which the actuator 120 is located on the upper torso of the user. Depending on the application, the cable-driven actuator 120 can be located on different parts of the body such as the lower back, hips, legs, or attached to a backpack or rucksack. Figures 43A, 43B, and 43C show further details of the actuator system placed on the user's upper torso and its integration into clothing. The actuator 120 may be covered by a fabric piece (as shown in Figure 43C), a soft material, or a rigid case. If covered by a fabric piece, this piece may include flexible boneless elements to maintain a shape around the motor without contacting this piece. Robot clothing, over or under clothing version

[0386] The robotic garment can be designed to be worn over standard clothing or workwear. Figures 43A, 43B, and 43C show one embodiment of the device that can be worn over clothing. In this embodiment, the system includes:

[0387] - Thigh Wrap , one on each leg. These pieces wrap around the user's thighs and can be worn over or under clothing.

[0388] - Load Balancing Element 212Connect both anchor members 110 and create a loop as previously described, creating an anchor point on the lower back.

[0389] - Shoulder strap The piece wraps around the user's shoulders and can be worn either over or under clothing.

[0390] - Actuation system : This component is designed to be integrated into a garment component. Figures 43A, 43B, and 43C show an example of how the actuator 120 is placed on the upper torso of the user. In this case, the cable drive system is fixed on one side to an anchor point created by a shoulder strap and on the other side to a click-in anchor point created by a load balance strap 112. As the actuator 120 moves, the cable can generate a force across the user's joints to provide assistance to the user. Other actuation concepts or mechanisms used to generate forces between the click-in anchor points can include fluid (pneumatic, hydraulic), electromechanical actuation, electrostatic, passive systems (springs, dampers, etc.), semi-active (clutches, etc.), etc. The actuator 120 can also be placed on other parts of the body such as the hips, thighs, backpacks, rucksacks, etc.

[0391] - Sensor Element In the proposed embodiment, the motion sensors 230 may be integrated, for example one per thigh to measure leg motion, one on the lower body and / or one on the upper body. The load sensor elements 230 may be used to measure or estimate the forces the device generates on the person. This may include current sensing, torque sensors, load cell sensors, strain gauges, sensors whose electrical properties change when stretched or loaded, etc. Other embodiments may include pressure sensors to measure ground reaction forces, force sensors on different parts of the body to estimate loads and user environment interaction forces, EMG sensors to estimate muscle activity, IMUs, gyros, accelerometers, encoders, resolvers, potentiometers, sensor units that change electrical properties (resistance, capacitance, etc.) based on deformations and loads applied to them, etc.

[0392] - Adjustment points 43A, 43B, and 43C have adjustment points that allow the length of the straps and the tension of the fabric pieces to be adjusted for different body sizes and user preferences. A ratchet system (e.g., BOA) or Velcro can be used to adjust the preload of the leg anchoring members 110, which can be adjusted based on thigh size and user preference / comfort. A ratchet system or Velcro piece in line with the load balancing straps 112 can be used to change the length of the load balancing straps 112. Changing the initial length of the load balancing straps 112 can help adjust the length of the device to different user heights or adjust the length of the cables based on the initial position of the thigh / wrap pieces. In the case of a passive system where the connection element 150 between the anchoring members 110 is a passive element 190, changing the length of this strap can set the initial tension of this piece, which can define parameters such as how much assistance the user receives as they move and the initial point in the user's movement where they start to generate force.

[0393] - Quick connect / disconnect feature The device may include quick connect / disconnect features to selectively attach and detach / disconnect the actuation, battery, or sensor components to the fabric components. These may include Velcro straps that wrap around and secure the sensor / actuator 120 features to the fabric components, click-in mechanisms, pogo pins, electrical connections that can electrically connect the actuation / power / sensor components with click-in for power / data transmission, etc. Quick connect / disconnect of the actuation, power, or sensor components may be practical to allow fabric elements that come into contact with the user to be washed or to allow the fabric or sensor portions of the wearable device 100 to be worn without being activated in some applications.

[0394] 44 and 45 show various users wearing the proposed system over their clothing to perform various tasks including reaching and lifting objects. Integration of anchoring members and connecting elements into clothing

[0395] Now, referring to FIG. 46A and FIG. 46B, in one embodiment, the components, structure, and actuator 120 system of the wearable device 100 can be incorporated into a typical garment for daily use. The garment can be defined as a standard outfit worn by a user for various uses (e.g., work clothes, worker uniforms, military uniforms, athletic clothing such as sports uniforms, casual clothing such as jeans, pants, T-shirts, or polo shirts). FIG. 46A and FIG. 46B show a version of the device that is comprised of work clothes with attachment points and a removable cable-driven sensor and actuator 120 system that can be clicked into the work clothes to assist the wearer. These components can be easily removed in just a few seconds by using components such as straps, Velcro straps, buckles, click-in mechanisms, etc., which is particularly important to facilitate the washing process of the fabric components. The user can then click the actuating and / or sensor components out of the fabric and wash the fabric components in contact with the user, or the wearer can wear the fabric components constantly and click the actuating and sensor components in as needed. The system is constructed of relatively non-stretch or low-stretch fabrics (e.g., woven fabric, webbing) that are attached (e.g., sewn, glued, top-mounted) directly to typical workwear to form the wearable device 100 structures (shoulder straps, thigh wrap, load balance straps 112) described above.

[0396] Also, if the pants and shirt are made of a relatively non-stretchy or low-stretch material (e.g., woven), the pant / shirt configuration may be part of the wearable device 100. As an example, if the pants are made of a woven material, a top attachment strap that can be used to keep the pants tight at a particular location (e.g., thigh) may be sufficient to transfer force to the thigh, eliminating the need to use a separate thigh wrap.

[0397] 47A, 47B, 47C, and 47D show further details of an embodiment of integrating a robotic part into a garment. In this embodiment, the actuator 120 is attached to the fabric part via a secure strap attachment, as shown in the top left of FIG. 47A. As an example, a relatively non-stretchable fabric strap is wrapped around a groove in a machined part of the actuator 120 and secured with Velcro or a buckle to hold the actuator 120 in place. In another embodiment, other approaches can be used, such as a click-in approach, where the actuator 120 integrates a click-in approach and the garment integrates a mounting plate 220 that the actuator 120 can click into, ensuring a mechanical and / or electrical connection with the fabric part. In this case, an electrical connector 222, such as a pogo pin, can be used to make the electrical connection. FIG. 48 shows an example where the actuator 120 is clicked into a mounting plate 220 sewn into a fabric part at the waist, where the mounting plate 220 has a click-in mechanism and an electrical mechanism. The actuation cable may click into one side of the load balance strap 112 via a mechanical connector and into the upper back via a connector or Velcro attachment, and in some embodiments, a conductive fabric may be used to power the sensors 230 (e.g., IMU, strain gauge, soft sensors 230) distributed in the fabric. FIG. 48 shows an example of powering the IMU sensors 230 on the upper back using conductive fabric and Velcro parts. Having a quick click into the mechanism can be useful in some cases, allowing the electromechanical parts to be removed from the garment to allow, for example, the garment to be washed. In other situations, this approach is useful when many users only wear the sensor-integrated fabric parts for part of the day and attach the electromechanical components as needed or to share these among different users.

[0398] Figures 47A and 47B also show further details of different fabric pieces that allow for adjustability and secure compression. The top left image (Figure 47A) shows an embodiment of a non-expanding fabric piece (webbing, sailcloth, woven fabric, etc.) that is sewn directly onto the fabric of a standard polo to make up a shoulder strap. Additionally, a ratchet system (e.g., BOA) is integrated into this strap to adjust the relative length of this shoulder strap to accommodate different users and provide customized comfort.

[0399] FIG. 47C and FIG. 47D show two implementations of integrating the thigh anchor members (e.g., thigh wraps) and load balance straps 112 directly into the workwear. FIG. 47C shows an example where the thigh anchor members 110 are sewn into the work pants and a ratchet system (e.g., BOA) can be used to apply compression to the integrated thigh anchor members 110 and ensure force transfer to the user's thighs. FIG. 47D shows an example where the thigh anchor members 110 are sewn inside the work pants and a ratchet system (e.g., BOA) can be used to adjust the compression of the thigh anchor members 110 components. FIG. 47C and FIG. 47D also show how the load balance straps 112 can be attached to the top of the anchor members 110 and run either over or under the pants, or through a path built into the pants (e.g., over a back pocket) and run from under the pants to over the pants. 47D also shows how an additional ratchet system (e.g., BOA) can be used to adjust the length of this load balancing strap 112 to accommodate different individuals or customize comfort features. A load balancing element 212 can then be clicked onto the lower end of the actuation cable to provide tension that is distributed to both thigh anchor members 110. Integrating sensors into clothing

[0400] 49, the sensors 230 can be designed to be fabric compatible to measure human motion, strap tension, interaction force / pressure and muscle strength. The proposed sensing wearable device 100 can use motion sensors 230 such as IMUs, accelerometers, gyros, etc. in combination with fabric compatible sensors 230 integrated into the system.

[0401] A) Highly stretchable sensor for motion tracking Measuring the kinematic and dynamic characteristics of joints as the wearer performs activities can be used for tasks such as recognizing various motions and providing assistance at the right time. The system may integrate customizable lightweight and stretchable electrostatic sensors 230 and conductive knitted fabrics as electrode and dielectric layers of silicone elastomers that can be integrated in the high stretch areas of the proposed wearable system. As shown in FIG. 49, in one embodiment, soft sensors 230 may be integrated to measure metrics such as body part motions such as torso bending (sensors along the back), torso twisting (sensors attached to the back and across the front of the garment), hip angle (sensors attached to the thighs), and to measure the user's posture and motion.

[0402] B) Low stretch sensor for tension measurement : The interaction force between the actuation cables is monitored and used to monitor the user interaction. Instead of a commercially available load cell, the system may include a soft tension sensor 230 that meets the requirement of sufficient sensitivity. This force / tension sensor 230 may be made insensitive to bending by encapsulating laser-cut conductors in an elastomeric material and bonding it to a rigid fabric. The robotic garment may integrate tension measurement sensors along the force path of the fabric (as shown in FIG. 49).

[0403] C) Fabric-integrated muscle activity sensor: EMG can be monitored using fabric electrodes laminated to the garment. These can provide information to the system to compare relative changes in peak / RMS muscle activity when performing task repetitions with different assistance patterns. We have demonstrated the feasibility of this approach by showing that muscle activity can be measured using individual electrodes as shown in FIG. 49. The sensor 230 can be integrated as an array of electrodes (suitably positioned to measure key muscles) on the inner surface of the wearable undergarment with encapsulated fabric-based signal traces. The electrodes are in a range that are firmly fixed to the wearer and can be due to material developments such as surface modification of fabric electrodes with conductive elastomers.

[0404] D) System Integration To integrate actuation and textile components, the system may leverage the inherent capabilities of knitted and woven materials. For example, the woven materials that make up the force pathways around the body are strong enough to support semi-rigid modules such as the mounting plate 220, robust mounting points for cables, and fixed attachment loops for tension sensors 230. Conversely, the knitted material of the garment can provide discreet, hidden pockets for pressure sensors 230, and low-friction, top-mounted channels for strain sensor 230 pathways. Load Control Wearable Device

[0405] For non-periodic tasks such as lifting, holding a stationary position, squatting, or raising one's arms to perform an overhead task, defining the assistance profile as a function of time or percentage of task completion may not be ideal. For such non-periodic tasks, it may be difficult to understand activities such as how long it takes a person to complete a lift or how long it takes to hold one's arms to complete the task. Furthermore, while performing the task, the wearer may decide to accelerate or change the direction of the motion, and thus the controller 240 providing assistance as a function of time or percentage of task completion may not adapt fast enough and may therefore feel disoriented. In this case, defining assistance as a function of the wearer's motion or posture has many advantages. It should be noted that providing load-based assistance to the user may also be useful for tasks that are periodic under some circumstances.

[0406] Here we present a control framework that solves these problems by calculating the tension the system applies as a function of one or more joint angles, velocities, or accelerations, or estimates of these variables. Joint or limb angles, velocities, and / or accelerations may be measured using sensor elements such as inertial measurement units, gyros, accelerometers, soft sensors 230 that change electrical properties in response to joint movement, encoders, potentiometers, etc. Because the desired force is a function of the wearer's motion and posture, we refer to this approach as load control.

[0407] As will be described in more detail below, embodiments of the present disclosure are directed to a wearable device 100 that includes:

[0408] At least one anchor member 110 configured for placement at a first body part of a person wearing the wearable device 100, at least one anchor member 110 configured for placement at a second body part of the person wearing the wearable device 100, and at least one connecting element 150 directly or indirectly connecting the at least one first body part anchor member 110 and the at least one second body part anchor member 110.

[0409] At least one sensor configured to measure information regarding one or more of the angle, velocity, and acceleration of one or more joints of the wearer spanned by the wearable device 100.

[0410] At least one controller 240 configured to detect the initiation and / or type of movement or posture of the wearer to be assisted by the wearable device 100, determine a desired tension to be generated within the wearable device 100 as a function of a given angle, velocity, or acceleration of one or more joints spanned by the wearable device 100 to assist the wearer in performing the movement or maintaining a static posture, and adjust the load on the wearable device 100 or the force provided by the wearable device 100 such that the desired tension is generated within the wearable device 100 at the given angle, velocity, or acceleration of the one or more joints.

[0411] The wearable device 100 may, in some embodiments, include at least one mechanism configured to selectively fix a length, change a length, or change a load of at least one connection element 150, where adjusting the load of the wearable device 100 includes fixing a length of at least one connection element 150 at a length configured to generate a desired tension in the wearable device 100 with a wearer's movement or posture at a given angle, velocity, or acceleration of one or more joints. The wearable device 100 may, in some embodiments, include at least two connection elements 150 and at least one mechanism 160 configured to selectively engage one or more of the at least two connection elements 150. Adjusting the load of the wearable device 100 includes engaging one or more of the at least two connection elements having a spring or damping constant, alone or in combination, and configured to generate a desired tension in the wearable device 100 with a wearer's movement or posture at a given angle, velocity, or acceleration of one or more joints. The wearable device 100, in some embodiments, may include at least one actuator 120, and adjusting the load of the wearable device 100 includes actively actuating at least one connection element 150 to generate a desired tension in the wearable device 100 at a given angle, velocity, or acceleration of one or more joints.

[0412] Controller 240, in various embodiments, may be configured to monitor information regarding one or more of the angles, velocities, and accelerations of one or more joints to detect when one or a combination of the angles, velocities, or accelerations exceed a threshold value indicating the start of an assisted movement or posture. Controller 240, in various embodiments, may be configured to monitor information regarding one or more of the angles, velocities, and accelerations of one or more joints to detect when one or a combination of the angles, velocities, or accelerations exceed a threshold value indicating the end of an assisted movement or posture.

[0413] Here we present a novel control framework that assists with a load-based approach. Note that while we discuss lifting and static posture holding, the approach may also be applied to other activities such as kneeling, walking, overhead tasks, grasping, etc. Figure 50 shows a sample control framework that consists of the following parts:

[0414] 1. The sensor elements are used to calculate joint angles, motions, or relative limb motions, and this information is used by the control algorithm to estimate the desired forces (e.g., onset, offset, assistance profile, etc.) that the device needs to provide to the user to assist with a particular task motion or posture; below is a sample embodiment of how the motion and posture information measured by the wearable sensors 230 is used to estimate the desired assistance. Additional sensor components may be added for various applications such as EMG, ground reaction forces, joint angles, body part / joint motions and positions.

[0415] 2. Assistance methods: Assistance profiles can be programmed as a function of measured body movements, postures and / or calculated or estimated from sensor components within the wearable device 100 (joint angles, body part movements and postures, differences in joint movements, sums of joint movements and other functions, EMG, ground reaction forces, etc.), and further, the load profile can differ for different states of the task.

[0416] 3. A method of transparency is used to prevent the system from applying forces when performing a task that it is not intended to assist with, or when the user / operator determines that assistance is not required. So in a transparent mode, the control algorithms can automatically detect when assistance is required, and the user or operator may have a manual input mechanism to turn assistance on / off. Sensor Element:

[0417] 51A, 51B and 51C, in one embodiment, inertial measurement units may be used to measure joint angles, velocities and accelerations. In this case, three IMUs are used, one on the upper back and one on each thigh, as best shown in FIG 51A.

[0418] During initialization, the user stands upright and the system measures joint angles to calculate offset values ​​in a standing posture (FIG. 51B). This procedure may not be necessary in some applications, but has benefits such as knowing the relative limb angles with respect to a standing posture. In another application, the system may use sensor inputs to estimate a neutral posture as the user performs a motion. For example, the controller 240 may monitor IMU measurements as the user performs a motion (walking, running, lifting, etc.) and use that information to define a neutral angle, for example, based on average signals from the motion sensor 230 to calculate a neutral posture, knowledge of the user's task and / or posture at a point in time, or the user pressing a button to initialize a neutral posture. The neutral posture may be updated over time by the controller 240 using this same method at various times throughout the day to compensate for small sensor position movements over time relative to the body and small fabric piece movements over time.

[0419] The system subtracts a standing offset value from the measured angle (Figure 51) so that an upright posture is defined as θ=0 for each IMU, and this angle increases as the user flexes their torso or thighs.

[0420] In this application example, the relative angle between the thigh and torso is used as the variable used to calculate the support profile; other relative angles (such as the relative angle between the torso and upper arm for shoulder support, the relative angle between the thigh and shin for knee support, the relative angle between the shin and leg for ankle support, etc.) may be used to define support profiles in other applications.

[0421] As shown in Figure 53A, the torso-thigh angle is defined as the relative angle between the average angle of the torso and both thighs as shown in the formula below. As an example, in a standing position, if the user has one leg in front and one leg behind at the same angle, the overall thigh angle will be neutral as shown in the formula below, but if the user has both legs at a positive angle, the overall thigh angle will be a positive value.

[0422] θΔ=θUB+0.5(θRT+θLT), where "Δ" means the torso-thigh relative angle, "UB" means "upper back", "RT" means "right thigh", and "LT" means "left thigh".

[0423] Figures 52A and 52B show this relative angle during walking, stair climbing, and lifting tasks.

[0424] The system, in this embodiment, may be configured to provide assistance during lifting tasks, but be transparent during other tasks that the system is not intended to assist with (e.g., walking, running, climbing stairs, driving a truck or car). As a result, the system needs to be able to distinguish between lifting or reaching tasks and walking, climbing, and running tasks. In some cases, the torso-thigh relative angle measured during a non-lifting task may appear to be the same as that measured during the early stages of a lifting task, when the user begins to bend downward from an upright position or begin to crouch. In particular, during walking and stair climbing tasks, the legs are typically out of phase, with one leg having a positive angle and the other a negative angle. These angles nearly cancel each other out in the above equation, resulting in smaller torso-thigh relative angle measurements during walking and stair climbing. During the early stages of a lifting task, the torso-thigh relative angle is similarly small, so the two angles may be confused. Without correction, the system may instead generate assistance during walking or stair climbing movements when it should be transparent; to avoid such problems, it may be necessary to configure the system to delay the onset of assistance until after the initial phase of the lifting movement, as shown in FIG. 53B.

[0425] The system can be configured to overcome this problem. In one embodiment, to better distinguish between lifting or reaching tasks and other tasks, a "penalty term" can be added to the equation for the relative torso-thigh angle, as shown below:

[0426] θ Δ =θ UB +0.5(θ RT +θ LT )-λ||θ RT -θ LT ||(λ>0), where “λ” denotes a penalty term for the difference between the right and left thigh angles.

[0427] In particular, the penalty term is configured to reduce small relative torso angles measured during walking and stair climbing, but does not affect small relative torso angle measurements taken during the initial phase of a lifting task. As shown in the figure, this penalty term has the advantage that the relative torso angle to the thigh is closer to neutral during walking and stair climbing, and the same angle is measured during a lifting task, which means that the penalty term does not affect the relative angle measurement during a lifting task, but can reduce the relative angle during tasks such as walking and stair climbing. Bringing the angle closer to neutral during non-lifting tasks is advantageous when using a threshold to start assisting the wearer, and as the angle approaches zero during non-lifting tasks, a smaller threshold may be used to distinguish between non-lifting and lifting tasks. FIG. 53B shows an example where an initial threshold in the displacement angle can be used so that the system does not provide assistance during the initial phase when the user bends the torso or leg forward. This threshold may be smaller when the penalty term compensates for the relative torso angle to the thigh during walking / running / stair climbing / etc. Alternatively, the algorithm may use motion sensor information to automatically set this initial offset value when worn by the user, hi another embodiment, the user may be able to modify this initial offset angle to suit specific preferences or needs.

[0428] In various embodiments, the assisted movement or posture may be a lifting movement or a squatting posture. In one embodiment, the controller 240 may determine a relative angle of one or more of the wearer's torso, thigh joint, and hip joint, and monitor the relative angle to detect when the relative angle exceeds a predetermined threshold indicative of the start of an assisted lifting movement or squatting posture. In another embodiment, the controller 240 may determine an average angle of the wearer's hip joint, and further monitor the average angle of the wearer's hip joint to detect when the average angle of the hip joint exceeds a predetermined threshold indicative of the start of an assisted lifting movement or squatting posture. The controller 240 may determine the average angle of the wearer's hip joint by calculating the relative angle between (i) the wearer's torso and (ii) the average angle of the wearer's thigh. In one embodiment, the controller 240 may be configured to identify whether the wearer is engaged in an activity other than an assisted lifting movement or squatting movement, and if so, may determine that any excess over the predetermined threshold is not indicative of the start of a lifting movement or squatting movement. Controller 240 may determine whether the wearer is engaged in an activity other than an assisted lifting or squatting by determining the difference between the angle of one hip joint and the angle of the other hip joint, applying a penalty term to the difference, subtracting the penalty difference from the average hip joint angle, and evaluating whether the resulting average hip joint angle determination exceeds a predetermined threshold. Controller 240, in one embodiment, may monitor the average hip joint angle of the wearer to detect when the average hip joint angle exceeds a predetermined threshold indicating the end of an assisted lifting or squatting position. Assistance Method - Programmable Load Assistance Control Structures

[0429] In this embodiment, the wearable device 100 may control the force delivered to the wearer. As shown in FIG. 54, the controller 240 consists of two main components, a high level controller and a low level controller, each generating output signals based on various sensor inputs and control logic. The high level incorporates active / invisible and assistive methods. Based on the signals of the IMU sensors 230 on the torso and legs, the high level controller 240 may classify the wearer's current movement / posture and decide whether the suit should be assisted or invisible, but alternatively, the user may press a button or other users may be involved to toggle the assistance on / off or put the system into the invisible mode. The output signal of this high level control block is the desired force command Fd that the garment should exert on the human body.

[0430] The difference FΔ between the actual applied force Fact measured or estimated by sensors (motor current, load cells, force sensors, torque sensors, etc.) and the desired force Fd generated by the high level controller 240 is used as input to the low level controller 240. The purpose of this block is to ensure stable dynamic tracking of the desired force by sending control signals to the actuators 120. This is achieved by using further input signals read from sensors 230 such as IMUs and motor encoders. Low Level Control

[0431] In one embodiment, the low level control scheme for controlling the forces or tensions in the system consists of two parts (Figure 55): the admittance controller 240 and the feedforward model. The motor controller 240 implements two cascaded PI loops for speed control and current control.

[0432] Admittance Control: A control structure that may be able to control the desired force on the wearer is the admittance control structure. The admittance controller 240 measures or estimates the force using the sensor 230 and modifies the trajectory of the robot due to the interaction force. In this embodiment, the admittance controller 240 is used to control the interaction force between the wearer and the wearable system. A general definition of the load is obtained from a damped simple harmonic oscillator. The second order differential equation that describes the dynamic behavior of this mechanical system is as follows:

[0433]

number

[0434]

number

[0435] sZ(s)=F(s) / X(s) Conversely, admittance is given as the inverse of the load (impedance) to define the dynamic relationship between output flow and input force.

[0436]

number

[0437] Y(s)=Z(s) -1 =(ms+d+k / s) -1

[0438] The admittance control principle used to control the interaction forces exerted on the human by the proposed active garment can be illustrated with a similar mechanical diagram, as shown on the right side of Fig. 17. Here, the spring is replaced by an actual force F_act, the force exerted by the human and parts of the wearable system on the actuator 120, which is measured by a load cell.

[0439] The virtual damping dv affects the steady-state value of the response, and the ratio of the virtual mass to the damper changes the dynamics of the system, so reducing dv makes the controller 240 more sensitive to force errors, but may also make the system less stable.

[0440] A way to address this problem is variable control, where parameters are adjusted online according to human intent or control task. In this task, a state machine operating at a higher level in the controller 240 is used to vary the virtual damping dv depending on different tasks. The desired virtual damping depends on the target force level (Fd) and dynamics (dFd / dt) and the dynamics of the wearer's relative torso-thigh angle (dθ / dt). Therefore, the appropriate virtual damping value for each condition can be defined as a function:

[0441]

number

[0442] The right plot of FIG. 56 shows a damping scheduling approach that qualitatively outlines the range of suitable damping values ​​for controlling the garment based on the dynamic coefficient η.

number

number

number

number

[0443] The tracking performance of the admittance controller 240 can be improved for soft garments by incorporating information about the garment, the wearer's motion, and the actuation system in the form of a feedforward model. Similar to those approaches, a motion and suit feedforward model was developed for the garment, as described next. Suit stiffness feedforward model

[0444] As mentioned before, when a load is applied to the suit, the system may deform. To model the suit's deformation, three male subjects were asked to stand in a leaning posture (torso tilt angle of approximately 60° in the sagittal plane). During the test, the cable was tensioned at 60 mm / s until a maximum force of 250 N was reached, and then released at the same speed. This cycle was repeated 10 times for each subject.

[0445] Figure 57 shows the motion of the suit including cable tensioning (loading) and unloading hysteresis. As the graph shows, increasing force from a low level requires more active cable motion than increasing force from a higher force level. This information can be incorporated into the controller 240 using a feedforward model to improve tracking performance.

[0446] The elasticity of the suit can be modeled by the following equation:

[0447] F = f(pc) = α(e βpc -1) In the low-level controller 240, the velocity command is the feedforward velocity, which is a function of the desired force command Fd.

number

[0448]

number

number

number

number

number

[0449] A more challenging task for the low-level controller 240, however, is ensuring good force tracking performance for active tasks. These tasks include movements such as crouching and walking, and are characterized by independent relative motion between the thighs and torso. To measure these movements, in this embodiment, inertial measurement units (IMUs) are placed on the upper thighs and upper back. The IMUs include accelerometers, gyroscopes, and magnetometers, and, if properly positioned, the data generated by these sensors 230 can be combined to measure the Euler angles of the transformation between the global coordinate frame and the local coordinate frame of the IMU.

[0450] The thighs and torso are connected at the top to a cable 150 via a load balancing strap 112. The length of the relatively inelastic webbing is fixed, but the length of the cable can be changed by the actuation system 120. Figure 58 shows the theoretical relationship between the torso and thigh movements and the corresponding cable lengths, ignoring suit deformations. The diagram on the left shows a user standing in an upright position with cable pretension. During the bend, the position of the thigh and the connection point connected to the thigh via the inelastic webbing stays in the same position. As a result, the bending of the torso and the corresponding stretch of the skin Δptrunk are related to the same stretch of the cable (gain: 1), however, this is not true for the thigh movement, as shown in the right diagram of Figure 58. The movement of the thigh on one side, corresponding to a change in the webbing on the same side Δpthigh, is only related to a change in the cable length of 0.5Δpthigh (gain: 0.5).

[0451] The movement Δpi of body i is assumed to be proportional to the change in tilt angle Δφi measured by the IMU.

[0452] Δpi~Δφi Here, the Δ operator denotes the change in value between the current value of the variable and the upright position. The tilt angle φ is defined to be positive with respect to the corresponding body flexion.

[0453] This assumption, combined with the characteristics obtained from Figure 58, leads to the following hypothesis:

[0454] Δlcable=μ·[Δφtrunk+0.5(Δφthigh,l+Δφthigh,r)]

[0455] The following experiments were conducted to test the assumptions and hypotheses. Three male and one female subjects were instructed to perform a full range of motion (ROM) forward bending (1), full flexion and extension of the right and left legs respectively (2 / 3), and a full range of motion squatting. All four tasks were performed at a speed of 20 bpm and were repeated 10 times each. During the tests, the admittance controller 240 maintained the cable at a pretension level of 15 N, and both the cable length and the torso and thigh tilt angles were recorded.

[0456] The feedforward model equation is defined as the time derivative of equation (3.12), which results in the following equation:

[0457]

number

[0458] It should be noted that although this embodiment describes an admittance control approach to control the force delivered to the user, other approaches can also be used. Possible approaches include direct force control, which compares the desired force against a force (current, force, etc.) measured or estimated via the sensor 230, with a closed loop approach to minimize this error and ensure high quality tracking. Alternatively, DC control can be used if a model between the measured current and the delivered force is known, and the system converts the desired force to a desired current and closes the current loop to ensure proper tracking. High Level Control

[0459] Previously, we defined a methodology for tracking the desired force profile Fcmd. Force commands are generated by the high-level controller 240, which includes two main methods: the active / transparent method and the assistive method. The former distinguishes between wearer motions where the system needs to be assistive (e.g., lifting) and tasks where the system only tracks the wearer motion without applying forces (e.g., walking). If the active / transparent method identifies an "active" task that the device is supposed to assist, the assistive method generates an assistance profile Fcmd that is sent to the low-level controller 240. How to support

[0460] Defining the force command in the time domain presents several challenges. First, the force command must be increased or decreased in time as the wearer starts or stops the movement, respectively. Second, the dynamics of the user's lift must be considered to ensure that the desired force level is reached. A time-series approach where Fcmd=Fcmd(t) is difficult to achieve for several reasons. If we assume that the user lifts at a relatively constant speed, this speed needs to be estimated at the beginning of the movement to determine the appropriate time frame for each force increase and decrease. However, we found that suitable metrics such as torso velocity and acceleration measured by the torso IMU show high variability in the early phase of the movement (~10%) and are in similar ranges for slow (2.5 s) and fast (1 s) lifts. Furthermore, it is not justified to assume a constant lifting speed, since the user will accelerate or decelerate the movement during the lift, resulting in a jerk in the force command.

[0461] Therefore, in one embodiment, Fcmd=Fcmd(θ Δ), where the force command depends on the displacement angle θΔ, which is a measure of the deviation of the torso and thigh angles from a standing position, as shown in FIG. 51C. Thus, a displacement angle of zero corresponds to a standing position. Now, referring to FIG. 60A, FIG. 60B, and FIG. 60C, in one embodiment, the support may be defined as a function of the torso-thigh angle, as shown in FIG. 60A and defined above.

[0462] In Fig. 60B we show an example of how the load is programmed as a function of joint motion. In this case, the load provided when the user moves down (blue line) is different from when the user is moving up during lifting (black line) and when the user is moving up but closer to standing (green line). These assistance curves can be customized for different people and tasks by using the optimization methods described above, having a user interface and / or by using the results and evaluation of biomechanical experiments or reference biomechanical data (such as the average of many users performing one task) to set the parameters of the assistance profile when the user is wearing the system. The basic principle of the assistance method is to define a force profile such that when the wearer leans to grasp an object, the system applies an increasing force as the user bends his torso down to help move it as the torso moves in the same direction as gravity. As soon as the wearer starts to lift the object and stretch his back, the assistance may eventually decrease as the wearer approaches an upright position due to the fact that the force may rise to an even higher level and must work against gravity to help lift the additional load.

[0463] In various embodiments, the controller 240 may be configured to monitor information regarding one or more of the angles, velocities, and accelerations of one or more joints to detect one or more states of the assisted motion. The controller 240 detects various transitional states between lowering, lifting, and maintaining a static posture by examining changes in the angles, velocities, and accelerations of the joints, as described in more detail below. As an example, FIG. 60C shows the biological lumbar moment as a function of the relative angles of the thighs and torso as the user lifts a 25 pound weight. As the user moves down to lift the load, the lumbar moment follows a lower load line, followed by a transition period to a higher load state as the user transitions from moving down to moving up, and finally follows a higher load line as the user moves upwards with the load against gravity. The assistance technique may use measurements of the relative angles of the thighs and torso to provide tension to the user that resembles the biological torque or uses a scaled version of the biological torque (e.g., user selected maximum assistance via input controller 240, optimization-based maximum torque for a given movement or initial settings).

[0464] FIG. 61A and FIG. 61B show how different states of lifting and different loads can be defined. As shown below the graph in FIG. 61B and in the circle in FIG. 61A, the user starts in a standing position (θA=0), and as the user moves down to grab the object, the user bends the torso and / or thighs (θA>0), and while the user is moving down, the assist force is determined by the load profile (e.g., yellow line in FIG. 61B). When the user changes from a downward movement to an upward movement, the system records the angle (θA=C) at which the user makes this transition, and from that point defines the load profile as a factor of the user's torso-thigh angle (green line in FIG. 61B). The plot shows three different cases where it decides to move upward at three different angles (~20°, ~50° and ~100°). From this point on, a high load transition is defined, and as the user is moving against gravity, when the transition crosses the red line, the force is held at a maximum level until the user approaches a near-upright position, at which point the force drops off to unassisted as the user finishes lifting.

[0465] Alternatively to this profile, one could define support that follows a linear stiffness (green line in the diagram) when moving upwards, as opposed to holding the force as the user moves upwards. Holding the force at a higher level has the advantage that more force is delivered to the user as they move upwards, and following a linear high load profile has the advantage of matching the shape of the biological lumbar moment more closely.

[0466] The support of a linear passive system is shown by the red line in FIG. 60B. For an ideal passive device, the force exerted by the exoskeleton increases when the user moves downwards and decreases similarly when the user moves upwards, meaning that the wearer stores energy in the passive springs and regains the same energy when lifting an object. Increasing the stiffness of the springs will make the suit more supportive, but may limit the wearer's movements because more energy must be stored in the springs to reach a certain posture. On the other hand, lower stiffness allows a free range of movement, but the suit may not be supportive enough because the force level is low for a given posture. This trade-off between support and limitation can be resolved with an active system, as mentioned above. The user's gravity compensation during torso flexion is designed and adjusted to ensure that the suit is supportive but not restrictive, allowing a full range of movement, and can be adapted to different users and movements using techniques such as online optimization and manual selection of loads. With an active system, a much higher level of force can be obtained when the user moves upwards, compared to a passive device. In this way, the system delivers a net energy to the assisted joints in one lift cycle. Assuming that the moment arm of the force delivered by the suit is constant, the net energy is proportional to the area enclosed by the hysteresis curve, as shown in the blue region of Figure 60B.

[0467] FIG. 60C shows the main advantage of the load control approach, applying support to different biological moments and lifting speeds. FIG. 60C shows the mean and standard deviation of the lumbo-sacral moments during a 25 lb lift, plotted in impedance space with pelvic angle. The moments were calculated with inverse dynamics for five lifts and one subject using motion capture data and force plates. The L5-S1 biological moment is shown to scale linearly with trunk flexion, as shown by the linear regression line. As explained in the load control approach above, the suit's support is automatically adapted to the biological torque. The plot in FIG. 60C shows two active support curves, one with the user bending at approximately a 50 degree angle and one with the user bending at a 100 degree angle. Similar to the biological moments, the maximum force level for the 50 degree curve is 50% of the amount of the 100 degree trunk flexion level. The fact that the force is increased or decreased based on the user's motion and posture also allows for automatic adaptation of the support to different lifting speeds. Within a certain angular window, the force is varied between two force levels, and the faster the user moves within this angular window, the faster the force increases. This approach ensures a smooth force profile for every lift, regardless of the user's lifting speed. Both benefits can be viewed by plotting the assistance curve in the time domain. Detection of various conditions during lifting

[0468] As mentioned above, the assistance profile can be a function of the different states the user is undergoing. Figure 61A shows a state machine that detects if the user is standing upright, moving down, holding a posture, or moving up against gravity. These transitions define different load profiles that can be programmed. These transitions are detected by measuring joint angles, velocities, and / or accelerations to determine what the user is doing.

[0469] The user may transition from standing to moving down, which can be detected using a small threshold and compared to the torso-thigh angle. As an example, one algorithm may use a threshold of 7 degrees, which has been experimentally found to be successful in identifying lifting / reaching postures versus other actions such as standing, walking, etc. Note that the threshold can be changed automatically or manually depending on the application. From this point, assistance is defined along the yellow line as the user moves down. At some point, the user may decide to hold the posture, for example to sort various objects in a logistics application, to inspect a surface or perform a task such as drilling, or to bend down to perform a procedure like a surgeon in an operating room. In such situations, the force is maintained at a defined level. The holding state can be detected by using a small threshold on the relative angular movement (e.g., angular velocity) of the thighs and torso when the user transitions from a moving (e.g., moving up / down) state to a holding state. The relative angle of the thighs and torso when the user enters the hold state is stored as the hold angle (θhold) and is compared to the measured torso and thigh angles while in the hold state to determine whether the user intends to start moving upwards / downwards. At this point, the user can hold the position, continue moving downwards, or continue moving upwards. If the user moves slightly downwards away from the hold position, the state changes to "move downwards" and the assistance continues to follow the yellow load line. If the user moves away from the hold position to a larger angle, the state changes to "move up", at which point the assistance follows the green line in impedance space. At this point, the user can hold the position again (state changes to "hold") or continue moving upwards until the angle approaches the upright position at which point the state changes to "stand".

[0470] Referring to FIG. 62, the plot shows measurements from the device where five different users performed a series of 50 lifting cycles and the system rendered different types of load profiles. In particular, the plot shows the mean and standard deviation of assistance across these subjects and lifting repetitions. The users were presented with four different assistance methods, demonstrating the device's ability to deliver different types of forces, as defined in impedance space. These tests showed that the device i) can deliver different types of load profiles to the user, ii) the device achieves high tracking performance of the load profile, and iii) the state of lifting is successfully detected and the force accurately follows the load profile without significant delay.

[0471] Figures 63A and 63B show measurements of applied force and displacement angle from the device following the proposed assistance method as the user performed a series of four lifts at different speeds and ranges, starting from an upright position. The device was able to successfully track the load profile and transmit forces to the user, demonstrating that the method is adaptable to various speeds and ranges of motion.

[0472] In various ...

Claims

1. A wearable device, comprising: A first anchor member configured to be placed on an upper body of a wearer wearing the wearable device; A second anchor member configured to be placed on a leg of a wearer wearing the wearable device; and at least one connecting element that directly or indirectly couples the first anchor member to the second anchor member and bridges a plurality of joints between the wearer's upper body and legs; at least one electromechanical actuator that rotates to move the connection elements to generate moments about the plurality of joints, thereby generating assistive forces on the wearable device; at least one sensor configured to measure motion information related to one or more of angles, velocities, and accelerations of the plurality of joints spanned by the wearable device; At least one controller, determining whether to assist the wearer in a lifting motion based, at least in part, on the measured motion information; determining a first assist force for assisting the wearer in a downward movement in a lifting movement as a first function of the movement information; determining a second assist force for assisting the wearer in an upward movement in a lifting motion as a second function of the motion information; and and adjusting a load of the wearable device or a force provided by at least one electromechanical actuator such that the first and second assistive forces are generated within the wearable device according to the measured motion information, the first function, and the second function when assisting the wearer with a lifting motion, wherein the second assistive force of the second function increases from the first assistive force of the first function. The controller configured as described above; The wearable device comprising:

2. 2. The wearable device of claim 1, The at least one electromechanical actuator is Electric motor, a rotating pulley element coupled to an output of the electric motor; and a flexible elongated element wrapped around the rotating pulley element; wherein one end or an intermediate portion of the flexible elongate element is directly or indirectly connected to a second anchor member; and wherein actuation of the electric motor causes the flexible elongated element to wind around the rotating pulley, pulling the first anchor member and the second anchor member toward each other. The wearable device.

3. A wearable device according to claim 2, comprising: a tension sensor that measures a pulling force on the flexible elongate element, wherein the at least one controller is configured to compare the first assistance force to be generated and the second assistance force to be generated to a measured tension and adjust a load on the wearable device or a force provided by the wearable device in response to the comparison; and a load balancing mechanism connecting the flexible elongate element to the second anchor member, the load balancing mechanism being configured to connect the second anchor members together via the load balancing mechanism and to balance the first support force and the second support force between the legs of the wearer; Further comprising: The wearable device.

4. A wearable device according to any one of claims 1 to 3, the at least one sensor measures a posture and / or motion of the one or more joints of the wearer between the first anchor member and the second anchor member, the sensor The wearable device includes at least one of an inertial measurement unit, a gyroscope, an accelerometer, a strain sensor that deforms and / or changes electrical characteristics depending on joint movement, and / or an encoder of the electromechanical actuator.

5. A wearable device according to any one of claims 1 to 3, The wearable device, wherein the at least one sensor includes a first inertial measurement unit configured to be positioned on the wearer's upper back, and second and third inertial measurement units configured to be positioned on the wearer's legs.

6. A wearable device according to any one of claims 1 to 5, The at least one controller: determining the first assist force and the second assist force based at least in part on motion information related to angles of the plurality of joints; and determining whether to generate the first and second assistive forces in the wearable device according to the measured motion information, the first function and the second function based at least in part on motion information related to velocities of the plurality of joints; The wearable device is configured as follows.

7. A wearable device according to claim 6, comprising: the velocities of the plurality of joints include angular velocities related to a comparison of an upper body angle and a leg angle measured by the at least one sensor; The wearable device.

8. A wearable device according to any one of claims 1 to 7, The wearable device, wherein the at least one controller is further configured to transition to a first function when the motion information indicates that the wearer has initiated a downward movement from an upright position, and to a second function when the motion information indicates that the wearer has initiated an upward movement to an upright position.

9. A wearable device according to any one of claims 1 to 8, The wearable device, wherein the at least one controller is configured to adjust the force provided by the load or the wearable device such that a second assistance force according to a second function increases to a higher level than a most recently applied first assistance force.

10. 2. The wearable device of claim 1, The at least one controller: determining a third assist force to be generated within the wearable device to assist the wearer in a transition between a downward movement in a lifting motion and an upward movement in a lifting motion, wherein the third assist force is determined as a third function of the measured motion information and is related to an assist force in a transition between the first function and the second function; and The wearable device, further configured to adjust a load on the wearable device or a force provided by the wearable device such that the third assistive force is generated in accordance with the measured motion information and the third function.

11. 2. The wearable device of claim 1, The at least one controller: The wearable device is further configured to adjust a load on or a force provided by the wearable device such that the first function increases the first assist force when the wearer moves downward during a lifting motion, and the second function applies a peak assist force after the wearer begins an upward motion during a lifting motion.

12. A wearable device according to any one of claims 1 to 11, The at least one controller: The wearable device is further configured to determine a first assist force and a second assist force based at least in part on motion information relating to angles of the plurality of joints, such that the second function provides an increased assist force from the first function at any angle of the plurality of joints during a lifting motion, such that the second function provides a stronger assist force when the wearer is moving upward than when the wearer is moving downward.

13. A wearable device according to any one of claims 1 to 12, The at least one controller: The wearable device, further configured to adjust a load on the wearable device or a force provided by the wearable device such that the second function applies a supporting force that decreases as the wearer approaches an upright position.

14. A wearable device according to any one of claims 1 to 13, The at least one controller: The wearable device is further configured to monitor motion information to detect a transition of the wearer to an upward motion phase or to a downward motion phase, wherein a first assist force is generated according to a first function during the downward motion phase, a second assist force is generated according to a second function during the downward motion phase, and an assist force related to both the first function and the second function is generated during a phase between the upward motion phase and the downward motion phase.

15. A wearable device according to any one of claims 1 to 14, comprising: The at least one controller: applying a clearing or pretensioning force when the measured motion information indicates the wearer is walking or climbing stairs; and The wearable device is configured to delay initiating generation of a first assistive force in accordance with the measured motion information and a first function until the measured motion information indicates that the wearer is no longer walking or climbing stairs.

16. A wearable device according to any one of claims 1 to 15, Each of the first function and the second function is Includes a support profile, and The wearable device further includes a user interface that enables a wearer to set parameters of the assistance profile when the wearer wears the wearable device, for customizing the assistance profile for different wearers and different tasks.

17. Measuring motion information relating to at least one of angles, velocities, and / or accelerations of a plurality of joints of the wearer spanned by the wearable device between the wearer's upper body and legs; determining whether a lifting motion of the wearer should be assisted by an electromechanical actuator of the wearable device based at least in part on the motion information; determining a first assist force to be generated in the wearable device as a first function of the motion information, the first assist force being adapted to assist the wearer in the downward movement of the lifting motion; determining a second assist force to be generated in the wearable device as a second function of the motion information, the second assist force assisting the wearer in the upward movement of the lifting motion; and adjusting a load on the wearable device or a force provided by the wearable device such that, when assisting a wearer with the lifting motion, the first and second assistive forces are generated within the wearable device according to the measured motion information, the first function, and the second function, wherein the second assistive force of the second function is increased from the first assistive force of the first function.

18. 18. A method according to claim 17, comprising the steps of: Measuring a pulling force within the wearable device; comparing the first assist force to be generated against the measured pulling force with the second assist force to be generated; adjusting a load on or a force provided by the wearable device according to the comparison; and The method further includes balancing a weighting of the first and second support forces between an upper body and legs of a wearer.

19. 19. A method according to claim 17 or 18, comprising: determining the first and second assist forces based at least in part on angle-related motion information of the plurality of joints; and determining whether to generate the first and second assisting forces according to the measured motion information, the first function and the second function based at least in part on motion information related to velocities of the plurality of joints; The method further comprises:

20. 20. The method of claim 17, The method, wherein the velocities of the plurality of joints include angular velocities related to a comparison of an upper body angle and a leg angle.

21. The method according to any one of claims 17 to 20, comprising the steps of: transitioning to the first function when the motion information indicates that the wearer has begun a downward motion from an upright position; and transitioning to a second function when the motion information indicates that the wearer has begun an upward motion to an upright position; The method further comprises:

22. The method according to any one of claims 17 to 21, The method further includes adjusting the load on the wearable device or the force provided by the wearable device so that a second assistance force according to a second function increases to a higher level than the most recently applied first assistance force.

23. The method according to any one of claims 17 to 23, comprising the steps of: determining a third assist force to be generated to assist the wearer in a transition between a downward movement of a lifting motion and an upward movement of a lifting motion, wherein the third assist force is determined as a third function of the measured motion information and is related to an assist force in a transition between the first function and the second function; and adjusting a load on a wearable device or a force provided by a wearable device such that the third assistive force is generated in accordance with the measured motion information and the third function; The method further comprises:

24. The method according to any one of claims 17 to 23, comprising the steps of: adjusting the load of or the force provided by the wearable device such that the first function increases the first assistive force as the wearer moves downward during a lifting motion, and such that the second function applies a peak assistive force after the wearer begins an upward motion during a lifting motion; The method comprising:

25. The method according to any one of claims 17 to 24, comprising the steps of: determining the first assisting force and the second assisting force based at least in part on motion information related to angles of the plurality of joints; and increasing the assist force provided by the second function from the first function such that the assist force is greater when the wearer is moving upward than when the wearer is moving downward at any angle of the plurality of joints during a lifting motion; The method comprising:

26. The method according to any one of claims 17 to 25, and adjusting the load on or the force provided by the wearable device such that the second function applies a supporting force that decreases as the wearer approaches an upright position.

27. The method according to any one of claims 17 to 26, comprising the steps of: monitoring the motion information to detect a transition to an upward motion phase or to a downward motion phase; generating a first assist force according to a first function during the downward motion phase; generating a second assist force according to a second function during the downward motion phase; and generating an assist force related to both the first function and the second function during a phase between the upward motion phase and the downward motion phase.

28. A wearable device comprising: A first anchor member configured to be placed on an upper body of a wearer wearing the wearable device; a second anchor member configured for placement on a leg of the wearer; and at least one connecting element that directly or indirectly couples the first anchor member to the second anchor member and bridges at least one joint between the wearer's upper body and legs; at least one electromechanical actuator configured to generate a moment about the at least one joint, thereby generating an assistive force on the wearable device; at least one sensor configured to measure motion information relating to one or more of angle, velocity, and / or acceleration of the one or more joints of the wearer spanned by the wearable device; At least one controller, determining whether to assist the wearer in bending and standing movements based, at least in part, on the measured movement information; determining a bending-to-stand assist force to be generated by the wearable device to assist the wearer in bending-to-stand as a first function of at least one angle of the motion information; determining a transitional assist force to be generated by the wearable device to assist the wearer in transitioning between bending and standing, the transitional assist force being applied when a velocity of the motion information is within a predetermined range; and and when it is determined that a movement in the bending-to-standing direction should be assisted, the controller adjusts the assist force provided in the wearable device according to the bending-to-standing assist force determination, wherein the bending-to-standing assist force is increased from the transition assist force.

29. 29. A wearable device according to claim 28, comprising: The wearable device, wherein the velocity of the motion information includes an angular velocity related to a comparison of an upper body angle and a leg angle measured by the at least one sensor.

30. A wearable device according to claim 28 or 29, comprising: The at least one controller: determining a stand-to-bend assist force generated by the wearable device to assist the wearer in transitioning from stand to bend as a second function of the at least one angle of the motion information, wherein the stand-to-bend assist force of the second function of the at least one angle of the motion information is reduced from the bend-to-stand assist force of the first function of the at least one angle of the motion information; The wearable device further comprising:

31. A wearable device according to any one of claims 28 to 30, comprising: The wearable device, wherein the wearer's movements between bending and standing include phases that assist the wearer in holding or carrying an item against gravity and phases that assist the wearer in moving the upper body in the same direction as gravity and / or in the opposite direction to gravity.

32. A wearable device according to any one of claims 28 to 31, comprising: The wearable device, wherein at least one controller is further configured to adjust the assist force to be generated within the wearable device according to a determination of a transitional assist force to assist a wearer in a holding position in one of bending and standing.

33. 33. A wearable device according to claim 32, comprising: The wearable device, wherein the adjustment of the support force in the holding posture between bending and standing maintains the support force at a determined level.

34. Measuring motion information relating to one or more of angles, velocities, and accelerations of one or more joints of the wearer that the wearable device spans between the wearer's upper body and legs; determining whether to assist the wearer in a movement between bending and standing based at least in part on the measured movement information; determining a bend-to-stand assist force to be generated by the wearable device to assist the wearer in bending-to-standing as a first function of at least one angle of the motion information; determining a transitional assist force to be generated by the wearable device to assist the wearer in transitioning between bending and standing, wherein the transitional assist force is applied when a velocity of the motion information is within a predetermined range; and a step of adjusting an assist force provided in the wearable device when it is determined that a movement in a bending-to-standing direction should be assisted, such that a bending-to-standing assist force is generated in the wearable device according to at least one angle of the measured movement information and a first function, wherein the bending-to-standing assist force is increased from the transition assist force according to the first function.

35. 35. The method of claim 34, The method, wherein the velocity of the motion information includes an angular velocity related to a comparison of an upper body angle and a leg angle.

36. 36. The method of claim 34 or 35, determining a stand to bend assist force to be generated within the wearable device to assist the wearer in a stand to bend motion as a second function of at least one angle of the motion information, wherein the stand to bend assist force of the second function of the at least one angle of the motion information is reduced from the bend to stand assist force of the first function of the at least one angle of the motion information.

37. 37. The method according to any one of claims 34 to 36, wherein the wearer's movements between bending and standing include phases that assist the wearer to hold or carry an item against gravity, the phases assisting the wearer to move their upper body in the same direction as gravity and / or in the opposite direction to gravity.

38. The method according to any one of claims 34 to 37, comprising the steps of: The method, wherein the assist force to be generated within the wearable device is further adjusted according to determining a transitional assist force to assist the wearer in a holding position in one of bending and standing.

39. 39. The method of claim 38, The method, wherein adjustment of the support force in the holding positions between bending and standing maintains the support force at a determined level.

Images