Powered lower limb assistive device with obstacle avoidance

Abstract

A powered lower limb assistive device includes a joint and a controller that modulates an angle of the joint according to reference joint angles of a baseline control scheme. The controller is configured to modify the reference joint angles at the joint in response to detection of an obstacle in front of the device to reduce the risk of the assistive device colliding with the obstacle during a swing portion of a gait cycle. The device is equipped with a distance sensor to detect the presence of a stair step or other obstacle and to measure the distance between the device and the obstacle to inform the controller when the reference joint angles should be modified.

Description

[0001] POWERED LOWER LIMB ASSISTIVE DEVICE WITH OBSTACLE AVOIDANCE

[0002] This invention was made with government support under HD094772 awarded by the National Institutes of Health. The government has certain rights in the invention.

[0003] TECHNICAL FIELD

[0004] This disclosure is related to powered lower limb assistive devices and, in particular to lower limb prostheses or exoskeletons equipped to avoid obstacles.

[0005] BACKGROUND

[0006] Passive prosthetic legs enable individuals with lower limb loss to perform common daily activities such as level walking. However, these devices require significant compensations (e.g., hip hiking, hip circumduction, and / or vaulting) from the user to achieve foot clearance in certain situations. A lack of precise control and proprioception contributes to a high rate of stumbles, trips, and falls for transfemoral prosthesis users. Passive devices also cannot provide the positive mechanical work necessary for activities such as ramp ascent and stair ascent, which are only possible with additional compensations from intact joints. The associated overuse of intact joints increases the risk of secondary musculoskeletal injuries in the amputee population. Emerging powered prostheses have demonstrated advantages over passive or quasi-passive devices by providing net-positive mechanical work and active control.

[0007] Powered prostheses typically replicate normative able-bodied joint kinematics and / or kinetics using sensor feedback analogous to biological proprioception (e.g., joint angles, limb angles, loads). While this control philosophy can improve prosthesis biomimicry and gait symmetry, these devices generally lack higher-level perception to adjust joint trajectories to risky foot placements on stairs (e.g., too close to the next stairstep) or sudden changes in walking surface height (e.g., due to a curb or obstacle). In fact, one study reported a substantially higher rate of falls with a commercially available powered knee prosthesis than with a mechanical knee, despite participants having higher balance confidence with the powered device. One possible explanation is that normative joint kinematics, while more efficient, naturally provide less foot clearance than compensatory motions (e.g., mean of 15 mm vs. 34.1 mm, respectively, during level walking), thus increasing the risk of stubbing in uncertain environments. In contrast, the combination of proprioception and vision allows able-bodied individuals to make subtle adjustments to swing knee kinematics and toe trajectories based on their foot position relative to stairsteps or obstacles. Proper foot placement can reduce the risk of stubs, trips, and falls on stairs, but this is especially difficult for prosthesis users who lack proprioception and volitional control over the prosthesis.

[0008] Although stumble recovery strategies have been proposed for powered prosthetic legs, recent studies have focused on actively avoiding such events. In one example, a phase variablebased position controller was used to traverse obstacles by using residual thigh motion of the user to control the prosthetic joint’s progression through nominal able-bodied walking kinematics. In particular, a user is able to freeze the prosthetic knee in a flexed position by holding their hips in an extended position, and then circumduct the leg over the obstacle. However, this method requires an unnatural compensation to cross an obstacle, and the maximum obstacle height is limited by the maximum knee flexion during level walking. In another example, thigh kinematics were used as features for machine learning to predict obstacle crossing intent after toe-off and subsequently adjust knee and ankle flexion. But obstacle detection sensitivity with this method was relatively low. In another example, crossing obstacles of variable heights was enabled by adjusting maximum knee flexion based on the time-integral of the user’s residual thigh motion. However, this approach requires the user’s hip to unnaturally hold the weight of the prosthesis in an extended position over time to get extra knee flexion, requiring a longer swing period for taller obstacles. In another example, stair ascent was enabled over different stairstep heights by adjusting prosthesis swing kinematics based on the user’s residual thigh motion. However, this method also requires specific hip motions to provide adequate toe clearance, and the controller’s dependence on thigh velocity and acceleration makes it susceptible to inertial sensor noise.

[0009] SUMMARY

[0010] Embodiments of a powered lower limb assistive device include a joint and a controller that modulates an angle of the joint according to reference joint angles of a baseline control scheme. The controller is configured to modify the reference joint angles at the joint in response to detection of an obstacle in front of the device to reduce the risk of the assistive device colliding with the obstacle during a swing portion of a gait cycle.

[0011] Embodiments of the powered lower limb assistive device include a forward-facing ultrasonic sensor and a controller. The sensor is configured to measure a stance distance between the device and an obstacle detected by the sensor, and the controller modulates at least one joint angle of the device based at least in part on the stance distance. The assistive device may include any one or more of the above-listed features or the below- listed features in any technically feasible combination:

[0012] - the baseline control scheme includes the reference joint angles as a function of gait phase;

[0013] - the reference joint angles are based on normative values;

[0014] - an amount of modification of the reference joint angles is based at least in part on a distance between the assistive device and the obstacle;

[0015] - an amount of modification of the reference joint angles is proportional to a difference that is equal to a safe distance threshold minus a distance between the assistive device and the obstacle, when the difference is positive;

[0016] - an amount of modification of the reference joint angles is independent from a height of the detected obstacle;

[0017] - an amount of modification of the reference joint angles is based at least in part on a height of the detected obstacle;

[0018] - the reference joint angles are modified only when a distance between the assistive device and the obstacle is less than a threshold safe distance;

[0019] - the reference joint angles are modified only when a distance between the assistive device and the obstacle is less than a distance between the assistive device and the obstacle during a previous gait cycle;

[0020] - the reference joint angles are not modified when a turn is detected;

[0021] - the reference joint angles are not modified when a heading direction changes by more than a threshold value within a gait cycle;

[0022] - the device includes a distance sensor that measures a distance between the assistive device and the obstacle;

[0023] - the device includes an ultrasonic sensor that measures a distance between the assistive device and the obstacle;

[0024] - the device includes a forward-facing ultrasonic sensor located at an ankle end of a lower leg member of the device;

[0025] - the device includes a distance sensor that measures a distance between the assistive device and the obstacle only during stance; - the device includes a distance sensor that measures a distance between the assistive device and the obstacle during stance and only when a global angle of a lower leg member of the device is greater than a threshold value;

[0026] - the device includes a distance sensor that measures and calculates a distance between the assistive device and the obstacle as an average of multiple distance measurements during stance while a global angle of a lower leg member of the device is within a threshold value of zero with respect to vertical;

[0027] - the j oint i s a knee j oint;

[0028] - the joint is an ankle joint;

[0029] - the controller modifies the reference joint angles by providing additional plantarflexion at an ankle joint of the device;

[0030] - the joint is a knee joint, and the device includes an ankle joint, wherein the controller modulates an angle of the ankle joint according to the baseline control scheme and is configured to modify the reference joint angles at both joints in response to detection of an obstacle in front of the device to reduce the risk of the assistive device colliding with the obstacle during the swing portion of the gait cycle;

[0031] - the device is a powered knee-ankle prosthesis including a knee joint, an ankle joint, an upper leg member and a lower leg member coupled at the knee joint, a foot member coupled with the lower leg member at the ankle joint, a first actuator operable to apply a knee torque at the knee joint based on commands from the controller, and a second actuator operable to apply an ankle torque at the ankle joint based on commands from the controller, the controller modulating an angle of the ankle joint according to the baseline control scheme and being configured to modify the reference j oint angles at both joints in response to detection of an obstacle in front of the device to reduce the risk of the assistive device colliding with the obstacle during the swing portion of the gait cycle;

[0032] - the joint angle is modulated based on a phase variable defined as a function of real-time thigh angle and thigh angle at heel strike over an entire range between sequential heel strike events;

[0033] - the controller modifies the reference joint angles based at least in part on a distance between the assistive device and the obstacle measured during the swing portion of the gait cycle; - the controller modifies the reference joint angles based at least in part on closed-loop feedback from a distance sensor measuring a distance between the assistive device and the obstacle during the swing portion of the gait cycle; and / or

[0034] - the controller modulates a knee joint angle and an ankle joint angle according to a baseline control scheme in which reference j oint angles are a function of gait phase, and the controller is configured to modify the reference joint angles at one or both joints when the stance distance is less than a threshold value.

[0035] BRIEF DESCRIPTION OF DRAWINGS

[0036] FIG. 1 includes a conceptual diagram and block diagram schematically illustrating operation of a powered knee-ankle prosthesis equipped with obstacle avoidance.

[0037] FIG. 2 schematically illustrates an example of a powered lower limb assistive device.

[0038] FIG. 3 includes plots of left and right toe trajectories of a single able-bodied person relative to a stair profde.

[0039] FIG. 4 illustrates a relationship between thigh angle 0thand gait phase < > with a gait cycle divided into three states (S1-S3).

[0040] FIG. 5 schematically illustrates transition criteria among the three states of FIG. 4.

[0041] FIG. 6 is a modified photographic image of user climbing stairs using a prosthesis consistent with that of FIG. 2;

[0042] FIG. 7 is a modified photographic image of user traversing an obstacle using the prosthesis of FIG. 6.

[0043] FIG. 8 compares the kinematics of a baseline control scheme (BL) to the kinematics of the able-bodied data upon which the kinematics models of the baseline controller are based.

[0044] FIG. 9 compares the kinematics of the baseline control scheme (BL) of FIG. 8 to the kinematics of a stub avoidance control scheme during stair ascent and during level walking with and without an obstacle.

[0045] FIG. 10 compares durations of the stance and swing portions of gait during leveling walking for two users with and without an obstacle being present.

[0046] FIG. 11 includes safe line confusion matrices constructed to evaluate the ability to induce stubbing conditions during stair ascent;

[0047] FIG. 12 includes controller confusion matrices constructed to evaluate the effectiveness of the stub avoidance controller during stair ascent. FIG. 13 is a comparison of stub rates between the baseline controller and the stub avoidance controller for two users when the prosthetic foot is placed in front of a safe line during stair ascent.

[0048] DESCRIPTION OF EMBODIMENTS

[0049] Described below is a powered lower limb assistive device equipped with a distance sensor and an accompanying control system that modifies nominal joint angles of the device for stub and obstacle avoidance while reducing the need for residual limb compensations by the user. FIG. 1 includes a conceptual diagram (top) and block diagram (bottom) schematically illustrating operation of a powered knee-ankle prosthesis 10 equipped with obstacle avoidance. As discussed further below, the device 10 may be equipped with only a single, inexpensive ultrasonic sensor 12 to sense a distance dsito objects or terrain forward of the user U and the device 10 during stance — i.e., while the prosthesis 10 is in contact with the ground. Based on distance measurements, a stub avoidance portion 14a of a device control scheme 14 may modify reference joint angles of a baseline control scheme 14b and enforce the modified joint angles. In the illustrated example, the stub avoidance portion of the control scheme 14 introduces extra knee flexion A0kneat a knee joint 16 of the device 10 and / or extra ankle flexion A0ankat an ankle joint 18 of the device 10 to the baseline portion 14b of the control scheme 14. The baseline portion 14b of the control scheme 14 may be a phase-based kinematic control scheme that controls joint angles 0ne, 0ankas afunction of gait phase to emulate normative joint kinematics, at least during the swing portion of a gait cycle, according to a continuous data-driven model of level walking and stair ascent (see, for example, S. Cheng, et al., “Modeling the transitional kinematics between variable-incline walking and stair climbing,” IEEE Trans. Med. Robot. Bionics, vol. 4, no. 3, pp. 840-851, 2022). Here, “as a function of gait phase” means that each reference joint angle of the baseline control scheme is related either directly or indirectly to gait phase. The baseline portion 14b of the control scheme 14 may be referred to interchangeably below as the baseline control scheme or baseline controller. Similarly, the stub avoidance portion 14a of the control scheme 14 may be referred to interchangeably below as the stub or obstacle avoidance control scheme or the stub or obstacle avoidance controller.

[0050] FIG. 2 schematically illustrates an example of a powered lower limb assistive device 10 including an upper leg member 20, a lower leg member 22, and a foot member 24. The assistive device 10 may be a powered knee-ankle prosthesis, as illustrated in FIG. 2, or it may be a powered exoskeleton having equivalent artificial knee and / or ankle joints interconnecting equivalent structural limb members configured for removable attachment to an intact lower limb (e.g., a powered knee brace). While described and illustrated below in the context of an exemplary knee-ankle assistive device, this disclosure is applicable to any prosthesis or exoskeleton having at least one joint, including, for example, any combination of a hip joint, a knee joint, and / or an ankle joint.

[0051] In the illustrated example, one end of the upper leg member 20 is adapted for attachment to the end of the residual portion of the upper leg of an above-knee amputee user. An opposite end of the upper leg member 20 and is coupled with the lower leg member 22 at the prosthetic knee joint 16. The kneejoint 16 is a rotational joint that provides rotational movement of the lower leg member 22 relative to the upper leg member 20 about a knee axis 26. The lower leg member 22 extends from the knee joint 16 to the prosthetic ankle joint 18 at which the foot member 24 is coupled with the lower leg member 22. The ankle joint 18 is also a rotational joint that provides rotational movement of the foot member 24 relative to the lower leg member 22 about an ankle axis 28. The foot member 24 may be adapted to accommodate a shoe to provide device protection and cushioning.

[0052] The device 10 includes a first actuator 30 (e.g., a motor) configured to provide a knee torque Tknein a rotational direction at the knee joint 16 and a second actuator 32 configured to provide an ankle torque Tankin a rotational direction at the ankle joint 18. In the example of FIG. 2, a positive knee torque Tknecorresponds to knee flexion, and a negative knee torque Tknecorresponds to knee extension. Similarly, a positive ankle torque rankcorresponds to dorsiflexion, and a negative ankle torque Tankcorresponds to plantarflexion. The first actuator 30 may be rigidly mounted along the upper leg member 20, and its rotational output may be converted to the torque rkneapplied at the knee joint 16 via a transmission member rigidly attached to the lower leg member 22. Or the first actuator 30 may have an integral transmission and be concentric with the knee joint 16 with a housing and rotor rigidly affixed to the respective upper leg member 20 and lower leg member 22, or vice versa. Similarly, the second actuator 32 may be rigidly mounted along the lower leg member 22 or at the ankle joint 18 with its rotational converted to the torque rankvia one or more transmission members mounted on the foot member 24 or integrated with the actuator 32. Other arrangements are possible to provide torque at one or more of the prosthetic joints.

[0053] The device 10 includes at least one controller 34 configured to store and employ at least one control scheme 14 according to which the controller operates each actuator 30, 32 via a motion controller 36 to provide the desired torque rkne, Tankat each joint 16, 18. In the case of a kinematic control scheme, the controller 34 may operate to provide a desired joint angle 0^ and joint velocity 0 at each joint 16, 18. The controller 34 may be programable and / or programmed with the control scheme(s) 14 and in communication with each actuator 30, 32 to control its respective output torque or joint angle / velocity. In the illustrated example, the controller 34 is in two-way communication with each actuator 30, 32 to receive one or more inputs from the actuators, such as a real-time encoder position which can be used to determine real-time joint angles and / or angular velocities at each joint 16, 18, among other parameters.

[0054] The controller 34 may receive additional information from a sensor system 38 to implement the control scheme(s) 14. The sensor system 38 includes one or more sensors that collect information pertinent to the prosthesis and / or its surrounding environment. The sensor system 38 includes a distance sensor 12 configured to measure a distance dsiduring prosthesis stance from a solid object O located forward of the user and the device 10. In FIG. 2, the distance sensor 12 is an ultrasonic sensor mounted along an ankle end of the lower leg member 22 above and proximate to the ankle joint 18 of the device 10. The ultrasonic sensor 12 emits sounds waves and can determine the distance dstbased on a time factor of reflections of the sound waves. The illustrated object O is an upward extending step.

[0055] In other embodiments, the distance sensor 12 employs radar, lidar, or an optical vision system in lieu of or in combination with an ultrasonic sensor to detect the presence of and the distance from an upcoming obstacle O. The sensor system 38 may include other sensors, such as one or more inertial measurement units (IMUs) to provide kinetic and kinematic information about the user, prosthesis members, prosthesis joints, etc. to the controller 34 to implement the control scheme(s) 14. The sensor system 38 may for example be configured to provide real-time measurements of global thigh angle 0thto the controller 34 to estimate the gait phase of the device 10 and / or user U during locomotion or, as shown in FIG. 1, to help identify a present task in which the user is engaged, such as level walking, incline or decline walking, stair ascent or descent, sitting, standing, etc.

[0056] The control scheme 14 may be or may include a kinematic control scheme configured to mimic able-bodied joint and limb movement during locomotion, an impedance control scheme configured to mimic able-bodied joint torque during locomotion, a hybrid control scheme employing kinematic control during one portion (e.g., the swing portion) of each gait cycle and impedance control during another portion (e.g., the stance portion) of each gait cycle. Other control schemes are contemplated, with the controller 34 switching between and / or modifying its control scheme(s) 14 in real-time as necessary based on information received from the sensor system 38. Where an impedance control scheme is employed, the reference joint angles of the baseline control scheme may be or may include the equilibrium angle 0eqof the governing impedance control relationship T = —K(G — 9eq) — B9, where K, B, and 6 are joint stiffness, damping, and joint velocity, respectively. Torque or joint angle control at the joints 16, 18 is not limited to any particular controller or number of controllers. Rather, one or more controllers are used to control joint torque according to each of one or more control schemes. While FIG. 2 illustrates only one controller 34 and associated control scheme 14, some implementations include multiple controllers. The device 10 may for example include an impedance controller with one or more impedance control schemes and a separate kinematic controller with one or more kinematic control schemes. A finite state machine (FSM) may be employed to switch between kinematic and impedance controllers based on detection of swing or stance portions of the gait cycle, or to switch between a baseline control scheme 14b and a stub avoidance control scheme 14a depending on the real-time value measured by a distance sensor 12. The device 10 may include dedicated controllers for each joint as well. The device 10 may further include additional unillustrated components, such as power sources for the actuators and controller, cables, brackets, etc.

[0057] The distance dstcan be used to help determine a risky stepping condition that may require modification of the baseline control scheme 14b using the stub avoidance control scheme 14a. To determine risky stepping conditions that require active stub avoidance during stair ascent (SA), nominal able-bodied kinematics during stair ascent may be simulated with different foot placements relative to a next stairstep. A “clear step” during stair ascent may be defined as a step in which a trajectory of the toe never intersects a stairstep — neither the first stairstep ahead of the user, nor the second stairstep ahead of the user for step-over-step gait — in a given stair profile. A “stub step” may be defined as a step in which the toe trajectory intersects a stairstep of the stair profile. An intersection with the first stairstep can occur during early swing or with the second stairstep during late swing in step-over-step stair ascent. The knee and ankle joints 16, 18 of the powered assistive device 10 may normally follow predefined kinematic patterns based on timeinvariant able-bodied trajectories as implemented by the controller 34. But the resulting toe trajectories do not change based on foot placement on the stair tread — i.e., while emulating normative stair ascent, a kinematic controller will normally implement a particular toe trajectory regardless of whether the toe is near the rear edge of the current stair tread, near the riser of the next stairstep, or at any other toe position of stance.

[0058] To determine a reference safe distance between the prosthetic foot 24 and the next stairstep to avoid stubbing, motion capture data may be plotted relative to a static stair profile with horizontal shifts toward the toe trajectories, as in FIG. 3. FIG. 3 includes plots of able-bodied left and right toe trajectories of a single able-bodied participant relative to a stair profile having an approximately 30° incline — i.e., a 14.6 cm riser height and a 29.5 cm tread depth. To determine when the toe trajectory will intersect a stairstep, a minimum horizontal distance between the toe trajectory and the stairstep edge when the trajectory is below the stairstep height can be calculated using where Tx(t) and Ty(t) are respective time-dependent x- and y-coordinates of the toe trajectory throughout swing. For the cthstairstep edge for c 6 { 1, 2, 3, 4}, Sxand Syrepresent the x- and y- coordinate, respectively.

[0059] Once the minimum distance dminis found, the x-coordinate can be shifted toward the toe trajectory (to the left in FIG. 3) by the same amount, which is shown as the dashed line in FIG. 3, to mimic the situation when the toe stubs the stair if the foot is placed less than the distance dminfrom the stair. By simulating across multiple trials of experimental data at the same incline with multiple different subjects, a mean safe distance to ascend the stairs without stubs can be determined. For the specific example of FIG. 3, the mean safe distance was determined to be dmin= 9.6 cm.

[0060] With reference again to the example of FIG. 1, the controller 34 may operate to apply an adjustment to the kinematics of the baseline control scheme 14b in accordance with the stub avoidance control scheme 14a based at least in part on the distance dstprovided by the distance sensor 12. The baseline controller 14b may enforce virtual kinematic constraints of the knee and ankle joints 16, 18 determined by previously developed kinematics models. Such models may be derived from averaged knee and ankle kinematics from an able-bodied dataset for walking and stair climbing and return the desired joint angle 9^ and velocity 9^ for joint j G {kne, ank}, which are continuous functions of task and phase variables / , <p. The task variable may contain the realtime inclination and forward speed along the ground or stairs based on information from the sensor system 38.

[0061] With reference to FIG. 4, a phase variable <p may be estimated from global thigh angle 0th. The phase variable <p has a range between 0 and 1, with end points of the range defined at successive repeated events of a single gait cycle. The gait phase at any moment during locomotion can be represented as fraction or a percentage between 0 and 1 and is an indicator — roughly equivalent to a time factor during a steady gait — of how far into the present gait cycle the monitored leg is. FIG. 4 illustrates a typical relationship between thigh angle and gait phase (p. The stance portion of the gait cycle begins at a heel-strike (HS) event and ends at a toe-off (TO) event. The swing portion of the gait cycle begins at TO and ends at the next HS. In the illustrated example, the gait cycle is defined between successive heel strikes.

[0062] In FIG. 4, the gait cycle is divided into three sections or states SI -S3. The first section SI is characterized by a monotonically decreasing thigh angle 0thand is defined entirely within stance between HS and maximum hip extension (MHE). The second section S2 is characterized by a monotonically increasing thigh angle 0thand is defined between MHE and maximum hip flexion (MHF). S2 thus includes the stance-to-swing transition at TO. The third section S3 is characterized by a monotonically decreasing thigh angle 0th and is defined between MHF and the next HS.

[0063] Phase variable definitions have previously been extended from level walking to inclined walking and stair climbing, but these definitions typically only consider the two monotonic sections SI and S2 between HS and MHF. These definitions typically ignore or feed-forward the descending section S3 between MHF and HS, which diminishes the user’s volition over gait progression. Others have addressed this problem by redefining the gait cycle to start and end at MHF rather than HS. But that approach is sensitive to accurate and timely MHF detection in real time. Instead, it is proposed here to define the extra descending section S3 between MHF and HS based on an estimate of the thigh angle at heel-strike 0t^s, as illustrated in FIG. 4. The phase variable can then be calculated as follows based on the state of the thigh trajectory: where 0t^sis thigh angle at heel-strike. The coordinates represent respective 0th / pairs at MHE and MHF. In pilot experiments, some participants tended to swing the prosthesis backward after toe-off, causing MHE to occur later than TO. To avoid the associated phase lag, the phase may be allowed to jump from the current value to the expected TO phaseTOusing the mean value from the able-bodied dataset. To ensure the phase is continuous, a second-order Butterworth low-pass filter (e.g., 200 rad / s natural frequency, 0.9 damping ratio) can be used to smooth the jump if MHE occurs later than TO, in which case the governing equation for S2 can be replaced with where 0™ is thigh angle at toe-off

[0064] A finite-state machine (FSM) may be used to select from and transition between the three sections SI -S3 depending on measurements and detection of critical gait events. However, thigh angle 0th is not always as clean and smooth as depicted in FIG. 4. For example, the relationship could have additional local extrema, which would result in extra monotonic sections. To address this possibility, the FSM can be configured to transition not only sequentially from SI to S2 to S3 to SI .. ., but also to permit backtracking of the state when false detections occur. An example of the logic for this is summarized in FIG. 5, based on real-time detection of MHF and MHE. As illustrated in the example of FIG. 5, the S1-S2 transition occurs if MHE is detected or if early TO is detected. Here, early TO is indicated in the absence of foot contact (FC = 0) and when gait phase < > exceeds a threshold value. The threshold value in FIG. 5 is a gait phase value of 0.3. While in S2, a transition back to SI is permitted if the thigh angle is found to decrease (0th< 0t^HE) while there is foot contact (FC = 1). If MHF is detected while in S2, the FSM transitions to S3. While in S3, the FSM can transition back to S2 if 0th> 0tkHFor can transition to SI once HS is detected (FC = 1).

[0065] Referring again to FIG. 1, the controller 34 tracks the reference joint angles provided by the virtual constraints of the baseline control scheme 14b using the motion controller 36, which may be a proportional-integral-derivative (PID) position controller. The motion controller 36 may command actuator torques according to where 07and 07are the measured joint angle and velocity, respectively, and kp, kl, and kdJare constants representing proportional, integral, and derivative gains. These gains can be determined iteratively during pilot trials to maximize the tracking performance of each joint without providing excessive torques. Larger proportional and integral gains may be required during stance relative to swing to support and push the user’s weight upward. The same set of gains was used for both level walking and stair ascent for both participants of a case study discussed below.

[0066] The controller 34 selectively modifies the desired joint angles from the baseline kinematic control scheme 14b, where the depicted “switch” 40 enables the kinematics modification of the stub avoidance control scheme 14a. The modification is based on measurements from the distance sensor 12, which are used to determine how much extra knee flexion A0kneand ankle plantarflexion A0ankare necessary to clear the detected step or obstacle. In one embodiment, the modified knee and ankle reference angles arecalculated based on the measured distance dstduring each stance phase as where 6^eand 0®"kare the respective knee and ankle angles of the baseline control scheme 14b and dsafeis a safe distance related to dmin, as discussed above. The amount of the additional flexion is directly proportional to dsafe— dstwhen this difference is positive and zero otherwise — i.e., the closer the assistive device is to the obstacle during stance, the greater the amount of additional flexion. The distance dsimay be defined as the minimum distance measured by the di stance sensor 12 during the most recent stance phase, although some manipulation of the distance measurement may be employed as discussed further below. A rectifier is depicted by (■) and returns its value only when that value is positive. Otherwise the rectifier returns zero — i.e., when dstis greater than or equal to the safe distance dsa(e, no additional flexion is added to the baseline joint angle.

[0067] In another embodiment, the modified knee and ankle reference angles ®dnk arecalculated based on a distance dswmeasured during swing phase (FIG. 1, upper right) as where dclis a target clearance distance between the sensor and obstacle. The amount of additional flexion is directly proportional to dci— dswwhen this difference is positive and zero otherwise — i.e., the closer the assistive device is to the obstacle during swing, the greater the amount of additional flexion. The distance dswmay be measured in real time to allow closed-loop changes in knee and ankle reference angles throughout swing.

[0068] The proportional gains k^neand kankcan be tuned at the highest expected inclination / obstacle height to ensure the extra flexion is enough to avoid stubbing when the distance between the obstacle and the toe end of the foot portion of the device is zero. During pilot trials, the proportional gains,kneand kank, were tuned to achieve the prosthetic knee joint's maximum flexion angle after measuring zero distance between the prosthetic toe and an obstacle, allowing clearance of the highest obstacle in the experiments discussed below. The same gains were used during stair ascent trials. It was also determined during pilot trials that ankle plantarflexion creates more ground clearance than dorsiflexion for both stair ascending and obstacle crossing. Accordingly, kankmay be set to a negative value to generate ankle plantarflexion.

[0069] T[ ] is a second-order Butterworth low-pass filter that is applied only to the extra flexion term. This filter may help reduce potential instability caused by the step response of the extra flexion, especially when the flexion amount is large — e.g, when the foot member 24 is placed very close to the obstacle / stairs such that dsafe» dst. The Butterworth filter may be selected to maintain the original slope of joint angles before and after the flexion occurs. The kinematics modification may be applied immediately after TO and may be disengaged at a threshold gait phase value that provides sufficient time for the lower leg member 22 to extend to the desired position to prepare for HS. In the disclosed study, the kinematics modification was in effect from TO until > 0.8.

[0070] For stair ascent, the safe distance dsafecan be determined as:

[0071] ^safe—^min T d + Tn, where dminis the minimum distance to avoid stubbing as described above, d is the fixed distance between the distance sensor 12 and the toe end of the foot portion 24 (FIG. 2), and m is a margin of safety to avoid stubs when the foot is placed near but behind a line defined by dmin. For the disclosed study, dmin= 9.6 cm, d = 15.24 cm, and m = 1 cm.

[0072] For obstacle crossing, dstmay be defined as the minimum distance measured by the distance sensor 12 during stance when the global angle 6t(FIG. 2) of the lower leg member 22 is greater than a threshold value that ensures the distance sensor is detecting an object above ground level. dsafecan be tuned to the user based on a comfortable step length to traverse an obstacle. For the disclosed study the threshold for lower leg member orientation was > —5°, and dsafe= 55 cm, although the manner of determining dstmay be modified to avoid false positive detections of obstacles in some scenarios.

[0073] Experimental Methods and Results

[0074] The efficacy of a powered knee-ankle prosthesis equipped with stub avoidance consistent with the above description was evaluated in a case study in which two amputee participants used the prosthesis during stair ascent with risky foot placement and while crossing an obstacle during level walking. The stub avoidance control scheme 14a reduced the stub rate during stair ascent by 89.95% relative to the baseline control scheme 14b without requiring the excessive user hip flexion that is typical with passive prosthetic devices. The stub avoidance control scheme 14a also demonstrated an 87.5% stub avoidance rate during obstacle crossing without requiring excessive user residual hip motions. Additionally, an offline simulation was performed using pre-recorded ultrasonic sensor data over a multi-activity circuit to demonstrate an exceptionally low falsepositive activation rate indicating the robustness of the obstacle avoidance function of the prosthesis and its controller in real-world scenarios. FIGS. 6 and 7 are modified photographic images (backgrounds removed) of an exemplary prosthesis 10 consistent with the schematic representation of FIG. 2 and used in the case study presented below. FIG. 6 shows a user wearing the prosthesis 10 during stair ascent, and FIG. 7 shows the user wearing the prosthesis 10 while traversing an obstacle O during level walking. The prosthesis 10 is equipped with low-impedance actuators 30, 32 capable of producing high torque at each joint 16, 18 via a custom 22: 1 single-stage stepped-planet compound planetary gear transmission. The actuators 30, 32 are powered by G-SOLO Twitter R80A / 80VDC drivers (Elmo Motion Control, Petah Tikva, Israel) and are capable of precise position control with enough torque / power to perform demanding activities such as stair ascent. The control and signal processing code runs on a myRIO 1900 (National Instruments, Austin, TX) at 500 Hz as the controller 34. The prosthesis power source is four lithium-polymer batteries connected in series. The sensor system 38 includes a 3DM-CX5-25 IMU (LORD Microstrain, Williston, VT) to measure the global orientation 0thof the users’ residual thigh. Motor positions were measured using E5 optical quadrature encoders (US Digital, Vancouver, WA) with velocities estimated using Savitzky-Golay differentiation. The prosthetic foot 24 is mounted under a 6-axis load cell (M3564F, Sunrise Instruments, Nanning, China) at the distal end of the ankle joint 18. For purposes of simplicity, the controller 34 in this study was not configured to calculate forward walking speed or terrain incline. Instead, forward speed was assumed to be a constant 1 m / s for both stair ascent and level walking, and inclination was matched to the experimental conditions — i.e., 26.5° for stair ascent and 0° for level walking.

[0075] The distance sensor 12 is an ultrasonic sensor (LV-MaxSonar-EZTM, MaxBotix, Brainerd, MN) attached to the lower leg member 22 approximately 6 cm above the ankle joint 18. The particular ultrasonic sensor used in this case costs only $34 USD, has dimensions of 19.9x22.1 x 15.5 mm (6.8 cm3), weighs 4.3 grams, consumes 11 milliwatts, and has sampling frequency up to 40 Hz (25 ms). In comparison, a state-of-the-art LiDAR unit (e.g., CamBoard pico flexx, PMD Technologies) costs $399 USD, has dimensions of 68x 17x7.35 mm (8.5 cm3), weighs 8 grams, and consumes 540-680 milliwatts with a sampling time greater than 30 ms. As another example, the LiDAR in the iPhone 13 Pro (Apple Inc., Cupertino, CA) depends on the light and environment conditions (e.g., weak in sensing transparent objects) and is sensitive to motion and is thus insufficient for the prosthetic leg use case. Computer vision using an RGB camera can provide distance measurements, but is subject to disadvantages similar to LiDAR — i.e., higher cost, size, weight, power consumption, sampling time — with even higher computational time. While the ultrasonic sensor may be preferred for these reasons, other embodiments achieve the benefits of other aspects of the disclosed assistive device, control schemes, and related methods with by employing other types of distance sensors, such as LiDAR, radar, and optical vision systems.

[0076] TABLE I summarizes participant anthropometries and other general information for the two study participants. MFCL refers to Medicare Functional Classification Level (i.e., K-level), and both participants were congenital amputees.

[0077] TABLE I

[0078] At the outset of each experiment, the powered prosthesis 10 was fit with proper alignment to each participant by a licensed prosthetist. Both participants had more than 5 hours of experience using the powered prosthesis prior to the experimental study disclosed here. Each participant was additionally trained to use the prosthesis as equipped with obstacle avoidance for at least 2 hours prior to data collection.

[0079] The experimental setup for stub avoidance during stair ascent is illustrated in FIG. 6. The minimum distance dminto avoid stubbing based on able-bodied toe trajectories was indicated using strips of tape on the stairs. A step in front of the dminline is referred to below as a “close” step and indicates a high stub risk during the subsequent swing. A step behind the dminline is referred to below as a “far” step and indicates a low stub risk during the subsequent swing. Experiments were conducted with and without the stub avoidance control scheme 14a being active to compare kinematics and stub rates of the prosthesis with and without the extra joint flexion. To minimize user bias, the participants were not informed whether the stub avoidance controller was turned on or off. Four conditions were considered, including: 1) close step with the stub avoidance controller active, 2) far step with the stub avoidance controller active, 3) close step with the stub avoidance controller inactive, and 4) far step with the stub avoidance controller inactive.

[0080] At the beginning of each trial, each participant was instructed to step either close or far on the stairs, and the actual placement of the foot member 24 relative to the dminline was recorded via side camera and manually marked during post-processing. A total of 48 trials were performed with each of the four conditions assigned randomly but in equal number. Each participant took a 2-5 minute break after every 8 trials. To minimize transients and obtain stair kinematics more representative of a steady-state activity, the participants started each trial with at least one walking stride before transitioning to stair ascent. Similarly, to avoid a sudden stop on the last stair step, each participant was instructed to transition and continue after the stair ascent for some walking strides. The stub avoidance controller was only used during stair ascent, and each participant walked at a self-selected pace. The stair ascent trials were video recorded for later processing to identify stub events.

[0081] The experimental setup for obstacle crossing is illustrated in FIG. 7 with the level walking path and obstacle O shown in plain lines for clarity. The obstacle O was a box with dimensions 60.3 x 12.1 x 17.1 cm placed in the center of the level walking path between parallel bars on transversely opposite sides. Strips of tape T were placed an average step length away from both sides of the obstacle to improve foot placement consistency while traversing the obstacle O. For each trial, each participant walked at least one stride with the prosthesis 10 before and after crossing the obstacle O, leading with their biological leg during the crossing. Participants performed 8 trials at two different obstacle heights (12.1 cm and 17.1 cm) and took a 2-5 minute break between the different obstacle heights.

[0082] FIG. 8 compares the kinematics of the unmodified baseline control scheme (BL) to the kinematics of the able-bodied data upon which the kinematics models of the baseline controller are based. The comparisons include the kinematic data for one of the participants (TF01) during stair ascent (top row) and during level walking (bottom row). The kinematic data for the other participant is omitted in FIG. 8 but is qualitatively similar to that of TF01. FIG. 8 also includes kinematic data collected with the participant’s everyday passive prosthesis. This data is represented by the third unlabeled curve in each of the relevant plots. The vertical dashed lines indicate the value of the phase variable at toe-off These are not labeled in the figure, but toe-off of the baseline controller lagged able-bodied toe-off during stair ascent and led the able-bodied toe-off during level walking. Passive prosthetic toe-off led both the baseline and able-bodied toe- off during stair ascent, while lagging behind both during level walking.

[0083] During stair ascent, the baseline controller kinematics generally follow the same patterns with similar magnitudes as the able-bodied data, except for a noticeable phase shift that is likely due to differences in thigh motion. In contrast, the passive prosthesis data noticeably deviates from both the baseline controller and the able-bodied data. The thigh kinematics with the passive prosthesis have a very large standard deviation (shaded region) before the mid-swing phase, and the ankle joint stays nearly constant during stair ascent, indicating a lack of push-off assistance.

[0084] During level walking, the mean kinematics of the baseline controller are mostly within one standard deviation of the able-bodied data for the knee and ankle joints, and the phase variable has a nearly linear trend with a slight lag during the early-to-mid stance. Although the thigh trajectory of with the passive prosthesis during level walking has a similar shape to that of the able-bodied and baseline kinematics, it demonstrates a higher thigh extension and a later toe-off, indicating a shorter and faster swing. Additionally, the passive ankle joint exhibits large pl antarfl exion (negative angle) during early stance and early swing when no angular displacement is expected.

[0085] FIG. 9 compares the kinematics of the unmodified baseline control scheme (BL) to the kinematics of the stub avoidance control scheme (SC) for participant TF01 during stair ascent (top row) and during level walking with and without the obstacle O (bottom row). Due to similarity in kinematics, the data from obstacle crossing at both obstacle heights are pooled together in FIG. 9. The vertical dashed lines indicate the percentage gait cycle at toe-off, showing that the stub avoidance controller initiated an earlier toe-off than the baseline controller while crossing the obstacle during level walking.

[0086] The thigh angle comparisons indicate minimal residual hip compensation during stub avoidance, with the stub avoidance controller resulting in a thigh trajectory that is largely within one standard deviation of the baseline controller, especially during stair ascent. The root-meansquare error (RMSE) of the stub avoidance thigh trajectory during stair ascent is 4.54° ± 1.65° for TF01 and 3.69° ± 1.62° for TF02. The sagittal plane thigh trajectory of the stub avoidance controller during obstacle crossing includes a double peak near maximum hip flexion (MHF) that slightly overshoots the baseline peak thigh angle by 4.15° for TF01. The same trend was present with TF02 with a smaller overshoot of 3.03°. The thigh trajectory RMSE during level walking obstacle avoidance is 5.34° ± 1.64° for TF01 and 4.55° ± 1.15° for TF02 when comparing level walking without obstacle and stub avoidance with the obstacle.

[0087] The knee angle comparisons in FIG. 9 indicate that the stub avoidance controller provided about 10° of extra flexion during stair ascent and more than 40° of extra flexion during obstacle crossing. Only a small amount of ankle flexion was added by the stub avoidance controller during stair ascent — mostly during mid-to-late swing, which may be due to the relatively large variation near the minimum peak. Approximately 10° of extra plantarflexion was provided during obstacle crossing. The joint angles of the baseline and stub avoidance controllers at heel-strike match well for both cases. On average, TF01 and TF02 respectively reach maximum knee flexion at 75.29 ± 1.68% and 73.96 ± 0.96% of the level walking gait cycle when the obstacle is present, compared to 71.69 ± 2.13% and 72.71 ± 3.33% when no obstacle is present.

[0088] FIG. 10 compares the durations of the stance and swing portions of gait during leveling walking for both participants with and without the obstacle being present. On average, the swing time for TF01 was 24.9 ms longer with the obstacle than without the obstacle. In contrast, the difference in swing time with and without the obstacle was negligible for TF02 at less than 0.4 ms. The presence of the obstacle increased the stance time for both participants, with an increase of 280.8 ms for TF01 and an increase of 79.2 ms for TF02.

[0089] FIGS. 11 and 12 include a series of confusion matrices tabulating stub rates during stair ascent. To calculate stub rates, any form of contact between the prosthesis and the stair or obstacle the prosthesis is attempting to clear is counted as a stub. The safe line confusion matrices of FIG. 11 were constructed to validate the design of the experiment — specifically, the ability to induce stubbing conditions during stair ascent — by evaluating the stub rate when stepping in front of vs. behind the safe line (c / min in FIG. 6) with only the baseline control scheme and no added flexion from the stub avoidance controller. As expected, there is a reasonably high rate of true positives (49.09% on average), meaning a stub occurred because the participant stepped in front of the safe line, with a similar rate of false positives (50.91% on average), meaning no stub occurred despite stepping in front of the safe line. There is also a high rate of true negatives (93.55% on average), meaning no stub occurred when stepping behind the safe line, and a low rate of false negatives (6.45% on average), meaning a stub occurred despite stepping far from the next stairstep. These results indicate that, by instructing participants to step in front of the safe line during stair ascent, the experimental protocol can induce a high stubbing risk for the purposes of evaluating the stub avoidance controller.

[0090] The controller confusion matrices of FIG. 12 were constructed to evaluate the effectiveness of the stub avoidance controller when stepping in front of the safe line with and without the stub avoidance modifications to the baseline controller. This protocol restricts the analysis to high risk conditions where the stub avoidance controller is most necessary, but does not permit analysis of undesirable false activations of the stub avoidance controller (which will be considered later). The matrices of FIG. 12 show a very high rate of true positives (94.92% on average), meaning the stub avoidance controller successfully avoided a stub, and a very low rate of false positives (5.08% on average), where the stub avoidance controller failed to prevent a stub. The rate of true negatives (49.09% on average), meaning a stub occurred as expected because the stub avoidance controller was off, matches the true positive rate for the safe line evaluation in FIG. 11, which indeed corresponds to the same experimental condition. Similarly, the rate of false negatives (50.91% on average), meaning no stub occurred despite the stub controller being off, matches the false positive rate for the safe line evaluation in FIG. 11.

[0091] FIG. 13 illustrates some of the particular results from FIGS. 11 and 12. In particular, FIG. 13 is a comparison of stub rates between the baseline controller and the stub avoidance controller for both participants when the prosthetic foot is placed in front of the safe line. On average, the stub avoidance controller reduced stub rates by about 90%, with a reduction of 87.6% for participant TF01 and a reduction of 92.3% for participant TF02.

[0092] During obstacle crossing, each participant successfully performed 16 obstacle crossings with the stub avoidance controller active, despite a 25% stub rate for participant TF02, who encountered stubs twice on each of the 4.75-inch and 6.75-inch obstacles. Participant TF01 had a 0% stub rate during obstacle crossing.

[0093] The robustness of the stub avoidance controller was evaluated in an offline simulation using pre-recorded ultrasonic data from a multi -activity circuit including sitting, standing, level and inclined walking, turning, stair ascending / descending, and transitions between these tasks in a relatively narrow lab environment, all without obstacles along the circuit. In other work prior to the above-described case study, TF01 was fitted with the same distance sensor-equipped prosthesis and was instructed to perform as many continuous laps of these activities as possible, providing 660 strides for analysis. As a simulation, that data was processed by the above-described control scheme to determine whether extra joint flexion would be falsely triggered by the stub avoidance control scheme during those activities. Because the stub avoidance controller can produce a continuous range of extra flexion depending on the distance measurements (including very small values), instances of extra flexion added by the stub avoidance controller were classified as false events only for extra knee flexion of greater than 3.56° and only for extra ankle flexion of greater than 2.33°. These values correspond to the minimum amount of additional flexion that is perceptible at each joint by an able-bodied human. The overall false-positive rate was 1.23%, and those false events produced on average 10.22° ± 5.20° of extra flexion at the knee joint. Extra ankle flexion is proportional to that of the knee by a scaling factor of -0.25). All occurrences of extra flexion, including false events, returned the knee joint angle back to the nominal baseline value by heel-strike.

[0094] Nonetheless, some variations in the manner in which the distance dst is determined have been developed to further reduced false positive triggering of the stub avoidance controller 14a in complex surrounding environments and during maneuvers other than level walking over an obstacle and stair ascent. One alternative for obtaining dstincludes use of a moving average filter. In this alternative, multiple distance measurements (e.g., a 10-sample window) may be obtained while the lower leg member 22 is within an angle threshold of being perpendicular to the ground (0(« 0) during each stance, with dstcalculated as the average of those measurements. In another example, a distance measurement history can be used to determine whether or not the participant is approaching a detected obstacle by comparing the distance measured during the preceding stride with the distance dtmeasured during the current stride. The controller 34 can be programmed to provide the additional flexion A0 only when: dsafe. In another example, to eliminate false positives that may be triggered by turning maneuvers, a turn can be identified based on roll, pitch, and yaw measurements from a thigh-mounted IMU (e.g., the same sensor measuring global thigh angle) within one stride. The distance measurement can be voided if the heading direction is changed by more than a threshold amount (e.g., 30°) within a stride. These techniques were simulated with the obstacle-crossing data of the abovedescribed study to ensure they would not invalidate the results of that study. The simulation confirmed that the extra flexion before and after these controller changes are comparable for each trial.

[0095] Additional features of the prosthesis may include the ability to estimate the height of an upcoming obstacle to proportionally modify the amount of additional flexion provided by the controller while traversing the obstacle, closed-loop distance sensing, and / or an alternative location for the distance sensor. For example, the controller may be configured to use a correlation between thigh angle during MHF and stair / ob stacl e height to estimate the height of an upcoming stair or other obstacle. Or the distance sensor may be or include a LiDAR sensor capable of evaluating physical features of an obstacle, such as height. The distance sensor may include a higher sampling rate than the maximum 40 Hz sampling rate of the ultrasonic sensor used in the above-described experiment. A higher sampling rate would be useful with the closed-loop distance sensing described above, where a distance dswbetween the sensor and obstacle (or stair) is continuously measured during the swing portion of gait as the device approaches and attempts to clear the obstacle. The controller can then modify the reference joint angles of the baseline controller in real time if additional flexion is required to provide the target clearance distance. Another improvement includes relocation of the distance sensor to a lower position along the prosthesis than that illustrated in the figures. The distance sensor may for example be mounted on the shoe or otherwise along the foot member 24. This would eliminate any need for the user to wear shorts or roll up their pant leg to avoid blocking the sensor.

[0096] It is to be understood that the foregoing description is of one or more embodiments of the invention. The invention is not limited to the particular embodiment(s) disclosed herein, but rather is defined solely by the claims below. Furthermore, the statements contained in the foregoing description relate to the disclosed embodiment(s) and are not to be construed as limitations on the scope of the invention or on the definition of terms used in the claims, except where a term or phrase is expressly defined above. Various other embodiments and various changes and modifications to the disclosed embodiment(s) will become apparent to those skilled in the art.

[0097] As used in this specification and claims, the terms “e.g.,” “for example,” “for instance,” “such as,” and “like,” and the verbs “comprising,” “having,” “including,” and their other verb forms, when used in conjunction with a listing of one or more components or other items, are each to be construed as open-ended, meaning that the listing is not to be considered as excluding other, additional components or items. Other terms are to be construed using their broadest reasonable meaning unless they are used in a context that requires a different interpretation.

Claims

CLAIMS1. A powered lower limb assistive device comprising a joint and a controller that modulates an angle of the joint according to reference joint angles of a baseline control scheme, wherein the controller is configured to modify the reference j oint angles at the joint in response to detection of an obstacle in front of the device to reduce the risk of the assistive device colliding with the obstacle during a swing portion of a gait cycle.

2. The assistive device of claim 1, wherein the baseline control scheme includes the reference joint angles as a function of gait phase.

3. The assistive device of claim 1 or 2, wherein the reference joint angles are based on normative values.

4. The assistive device of any of claims 1 to 3, wherein an amount of modification of the reference joint angles is based at least in part on a distance between the assistive device and the obstacle.

5. The assistive device of any of claims 1 to 4, wherein an amount of modification of the reference joint angles is proportional to a difference that is equal to a safe distance threshold minus a distance between the assistive device and the obstacle, when the difference is positive.

6. The assistive device of any of claims 1 to 5, wherein an amount of modification of the reference joint angles is independent from a height of the detected obstacle.

7. The assistive device of any of claims 1 to 6, wherein an amount of modification of the reference joint angles is based at least in part on a height of the detected obstacle.

8. The assistive device of any of claims 1 to 7, wherein the reference joint angles are modified only when a distance between the assistive device and the obstacle is less than a threshold safe distance.

9. The assistive device of any of claims 1 to 8, wherein the reference joint angles are modified only when a distance between the assistive device and the obstacle is less than a distance between the assistive device and the obstacle during a previous gait cycle.

10. The assistive device of any of claims 1 to 9, wherein the reference j oint angles are not modified when a turn is detected.

11. The assistive device of any of claims 1 to 10, wherein the reference joint angles are not modified when a heading direction changes by more than a threshold value within a gait cycle.

12. The assistive device of any of claims 1 to 11, further comprising a distance sensor that measures a distance between the assistive device and the obstacle.

13. The assistive device of claim 12, wherein the distance sensor is an ultrasonic sensor.

14. The assistive device of claim 13, wherein the ultrasonic sensor is forward-facing and located at an ankle end of a lower leg member of the device.

15. The assistive device of any of claims 12 to 14, wherein said distance is measured only during stance.

16. The assistive device of any of claims 12 to 15, wherein said distance is measured during stance and only when a global angle of a lower leg member of the device is greater than a threshold value.

17. The assistive device of any of claims 12 to 16, wherein said distance is calculated as an average of multiple distance measurements during stance while a global angle of a lower leg member of the device is within a threshold value of zero with respect to vertical.

18. The assistive device of any of claims 1 to 17, wherein the joint is a knee joint.

19. The assistive device of any of claims 1 to 18, wherein the joint is an ankle joint.

20. The assistive device of claim 19, wherein the controller modifies the reference joint angles by providing additional plantarflexion.

21. The assistive device of any of claims 1 to 20, wherein the joint is a knee joint, the device further comprising an ankle joint, wherein the controller modulates an angle of the ankle joint according to the baseline control scheme and is configured to modify the reference joint angles at both joints in response to detection of an obstacle in front of the device to reduce the risk of the assistive device colliding with the obstacle during the swing portion of the gait cycle.

22. A powered knee-ankle prosthesis according to claim 21, further comprising: an upper leg member and a lower leg member coupled at the knee joint; a foot member coupled with the lower leg member at the ankle joint; a first actuator operable to apply a knee torque at the knee joint based on commands from the controller; and a second actuator operable to apply an ankle torque at the ankle joint based on commands from the controller.

23. The assistive device of any of claims 1 to 22, wherein the joint angle is modulated based on a phase variable defined as a function of real-time thigh angle and thigh angle at heel-strike over an entire range between sequential heel-strike events.

24. The assistive device of any of claims 1 to 23, wherein the controller modifies the reference joint angles based at least in part on a distance between the assistive device and the obstacle measured during the swing portion of the gait cycle.

25. The assistive device of any of claims 1 to 24, wherein the controller modifies the reference joint angles based at least in part on closed-loop feedback from a distance sensor measuring a distance between the assistive device and the obstacle during the swing portion of the gait cycle.

26. A powered lower limb assistive device comprising a forward-facing ultrasonic sensor configured to measure a stance distance between the device and an obstacle detected by the sensor,the device further comprising a controller that modulates at least one joint angle of the device based at least in part on the stance distance.

27. The assistive device of claim 26, wherein the controller modulates a knee joint angle and an ankle joint angle according to a baseline control scheme in which reference joint angles are a function of gait phase, wherein the controller is configured to modify the reference joint angles at one or both joints when the stance distance is less than a threshold value.

Images