Load reduction device, load reduction method, and storage medium storing program
By determining the user's repeated movements and using machine learning technology, comparing the extracted features and reference models, controlling the torque output of the load-reducing equipment, the problem that existing equipment is difficult to adapt to different motion modes is solved, and a more appropriate load-reducing effect is achieved .
Patent Information
- Application Number
- CN201980074217.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2019-08-21
- Filing Date
- 2019-11-14
- Publication Date
- 2024-11-12
- Estimated Expiration
- 2039-11-14
AI Technical Summary
Existing load reduction equipment is difficult to more appropriately reduce load for user movements, and the torque output is not sufficient to meet the needs of different motion modes.
By determining the user's repetitive motion, using machine learning to extract features and compare with the reference model, the torque output by the drive mechanism is controlled to reduce the load at the user's legs.
It achieves a more appropriate load reduction for user movement, adapts to different motion modes, and improves the efficiency and effect of load reduction equipment.
Smart Images

Figure CN112996636B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a load reduction device, a load reduction method, and a storage medium for storing a program. Background Art
[0002] A load reduction device is known which, when a user wears it, performs assistance of a load such as a walking motion of the user and reduces the load of luggage carried by the user. When a person can wear it, the load reduction device is sometimes called a power suit.
[0003] Patent document 1 discloses a technology that uses the output of at least one sensor to identify at least one predetermined gait event, adjusts the actuation curve of at least one actuator, and the actuation curve of at least one actuator generates a beneficial torque about at least one joint.
[0004] Patent document 2 discloses a technology that senses the movement of at least one joint or measures the movement of at least one joint, determines the movement state of at least one joint based on the sensed or measured movement of at least one joint, and controls a walking assistance robot based on the determination result about the operating state of at least one joint.
[0005] Patent Document 3 discloses a technology that acquires a state signal indicating the motion of an upper limb of a user and determines an assist timing, which is a timing for assisting the user's action, from a change in the state signal.
[0006] Patent document 4 discloses a technology for determining the driving amount of the actuator according to the sensor value when the pedestrian walks, based on a walking algorithm for determining the driving amount of the actuator driving the joint of the walking support device. Patent document 4 discloses that the walking algorithm is information generated by statistically processing walking data showing walking characteristics.
[0007] Citation List
[0008] Patent Literature
[0009] [Patent Document 1] Japanese Translation of PCT International Publication No. 2016-539723
[0010] [Patent Document 2] Japanese Unexamined Patent Application Publication No. 2015-089510
[0011] [Patent Document 3] Japanese Unexamined Patent Application Publication No. 2015-177863
[0012] [Patent Document 4] WO 2016 / 038824 Summary of the invention
[0013] Problems to be solved by the present invention
[0014] Although the torque output by the load reducing device required to reduce the load varies sequentially according to the momentum of the user's motion, it is desirable to perform more appropriate load reduction for the user's motion.
[0015] Therefore, an exemplary object of the present invention is to provide a load reduction device, a load reduction method, and a storage medium for storing a program therein that can solve the above-mentioned problems.
[0016] Means of solving the problem
[0017] According to a first aspect of the present invention, a load reduction device includes: a determination unit configured to determine a repetitive motion repeated by a user; and a torque control unit configured to control the torque output by a drive mechanism by comparing features extracted by machine learning based on the repetitive motion with a reference model to reduce the load on the user at the joints of the user's legs.
[0018] According to a second aspect of the invention, a load reduction method comprises: determining a repetitive motion repeated by a user; and controlling the torque output by a drive mechanism to reduce the load on the user at the joints of the user's legs by comparing features extracted by machine learning based on the repetitive motion with a reference model.
[0019] According to a third aspect of the invention, a program stored in a storage medium causes a computer of a load reduction device to perform processing, the processing comprising: determining a repetitive action repeated by a user; and controlling the torque output by a drive mechanism to reduce the load on the user at the joints of the user's legs by comparing features extracted by machine learning based on repetitive motions with a reference model.
[0020] Advantageous Effects of the Invention
[0021] According to the present invention, it is possible to provide a load reducing device capable of reducing the load more appropriately for the user's exercise. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] Figure 1 is a perspective view showing the configuration of a power suit according to an embodiment of the present invention.
[0023] Figure 2 is a diagram showing a hardware configuration of a control device according to an embodiment of the present invention.
[0024] Figure 3 is a functional block diagram of a control device according to an embodiment of the present invention.
[0025] Figure 4 is an operation block diagram showing the operation of the control device according to the first embodiment.
[0026] Figure 5 is a graph for explaining a method of determining a gait cycle according to an embodiment of the present invention.
[0027] Figure 6 : is a diagram showing an overview of a torque control process according to an embodiment of the present invention.
[0028] Fig. 7A is a graph showing an example of each feature in a gait cycle according to an embodiment of the present invention.
[0029] Figure 7B is a graph showing an example of each feature in a gait cycle according to an embodiment of the present invention.
[0030] Figure 8 is a flowchart illustrating the processing of the power suit according to the first embodiment.
[0031] Fig. 9 is an operation block diagram showing the operation of the control device according to the second embodiment.
[0032] Fig.10 is a flowchart illustrating the processing of the power suit according to the second embodiment.
[0033] Fig.11 is a graph showing the relationship between the output of the torque control unit and the target torque when the control according to the second embodiment is not performed.
[0034] Fig.12 is a graph showing the relationship between the output of the torque control unit and the target torque when the control according to the second embodiment is performed.
[0035] Fig.13 is a diagram showing a minimum configuration of a control device according to an embodiment of the present invention. DETAILED DESCRIPTION
[0036] Hereinafter, a load reducing apparatus, a load reducing method, and a storage medium for storing a program therein according to an embodiment of the present invention will be described with reference to the accompanying drawings.
[0037] Figure 1 is a perspective view showing the configuration of the power suit according to the present embodiment.
[0038] The power suit 100 is one aspect of the load reduction device. The power suit 100 is composed of a frame portion 11, a belt 12, a hip actuator 13, a knee actuator 14, an ankle actuator 15, a shoe sole plate 16, a foot strap 17, a shoe sole load sensor 18, a foot sole load sensor 19, a loading platform 20, a control device 21, a battery 22, a hip joint sensor 23, a knee joint sensor 24, an ankle joint sensor 25, and an acceleration sensor 26. As an example, the frame portion 11 is roughly divided into a first frame portion 111, a second frame portion 112, and a third frame portion 113.
[0039] like Figure 1 As shown, the power suit 100 is configured as follows to support a loading platform 20, which is one aspect of a mechanism for supporting luggage as an example. That is, the power suit 100 has a first frame portion 111, and the left and right hip actuators 13 are rotatably coupled to the first frame portion 111 and the second frame portion 112, which correspond to the left or right thigh portion of the user wearing the power suit 100, respectively. The left and right knee actuators 14 are rotatably coupled to the corresponding second frame portion 112 on the left or right side of the user wearing the power suit 100, and are rotatably coupled to the corresponding third frame portion 113 along the left or right calf portion of the user wearing the power suit 100. The ankle joint actuator 15 is rotatably coupled to the corresponding third frame portion 113 on the left or right side, and the corresponding sole plate 16 is provided on the back of the foot strap 17 on the left or right side of the user wearing the power suit 100. The actuators 13 , 14 , and 15 are driving mechanisms that output a torque that relieves a load applied to the user at each joint of each leg of the user.
[0040] The user wearing the power suit 100 attaches his / her left and right feet to the corresponding foot straps 17, and fixes the first frame portion 111 to the waist using the belt 12 so that the first frame portion 111 is closely attached to the waist. The power suit 100 has a structure that contacts the sole of the foot via the frame portion 11 and the hip actuator 13, the knee actuator 14, and the ankle actuator 15, thereby releasing most of the load of the luggage and the load of the power suit 100 to the ground. The user turns on the control device 21 of the power suit 100. The control device 21 controls the hip actuator 13, the knee actuator 14, and the ankle actuator 15 to transfer as much of the device weight as possible to the walking surface via the frame portion 11 and the hip actuator 13, the knee actuator 14, and the ankle actuator 15, which is the sum of the load of the luggage loaded on the loading platform 20 and the weight of the power suit 100. Thus, the power suit 100 reduces a burden, such as the load of luggage of a user who wears the power suit 100 , loads the luggage on the loading platform 20 , and walks.
[0041] The hip joint sensor 23 is installed in the hip actuator 13, and detects the hip joint angle, that is, the angle formed between the first skeleton part 111 and the second skeleton part 112, through an encoder. The knee joint sensor 24 is installed in the knee actuator 14, and detects the knee joint angle, that is, the angle between the second skeleton part 112 and the third skeleton part 113, through an encoder. The ankle joint sensor 25 is installed in the ankle actuator 15, and detects the ankle joint angle, that is, the angle between the third skeleton part 113 and the sole plate 16, through an encoder. The hip joint sensor 23, the knee joint sensor 24, and the ankle joint sensor 25 detect the angle of each joint of each leg of the user (hereinafter referred to as "joint angle"). The acceleration sensor 26 detects the acceleration of the user's lower limbs.
[0042] The sole load sensor 18 is provided at the bottom of the foot strap 17, corresponding to the shoes worn by the user. Moreover, the sole load sensor 18 is provided on the ground contact surface side of the sole plate 16 and the foot strap 17 to cover the entire back side of the sole plate 16 and the foot strap 17, the sole plate 16 transfers the weight of the power suit 100 and the luggage to the ground contact surface, and the foot strap 17 transfers the weight of the user to the ground contact surface. The sole load sensor 19 is provided in the foot strap 17 to cover the entire sole surface so that the weight of the user can be measured from the sole of the foot. For example, the sole load sensor 19 may be provided between the insole of the foot strap 17 and the sole plate 16, or may be provided on the upper part of the sole plate 16.
[0043] As an example, the shoe sole load sensor 18 and the foot sole load sensor 19 are sensors in which electrodes are arranged in a matrix form on the front and back surfaces of a thin sheet-like insulator. The shoe sole load sensor 18 and the foot sole load sensor 19 measure the resistance of the grid points of the electrodes and output the measured values to the control device 21. The control device 21 calculates the pressure applied to each grid point and the load on the entire surface of the sensor sheet based on the resistance value of each grid point.
[0044] Figure 2 is a diagram showing the hardware configuration of a control device.
[0045] As shown in the figure, the control device 21 is a computer equipped with hardware such as a CPU (Central Processing Unit) 101 , a ROM (Read Only Memory) 102 , a RAM (Random Access Memory) 103 , a signal input / output device 104 , and a wireless communication device 105 .
[0046] The signal input / output device 104 inputs signals output from the sole load sensor 18, the sole load sensor 19, the hip joint sensor 23, the knee joint sensor 24, the ankle joint sensor 25, and the acceleration sensor 26. The signal input / output device 104 outputs control signals for controlling the hip actuator 13, the knee actuator 14, and the ankle actuator 15. The control device 21 operates by power supplied from the battery 22.
[0047] The wireless communication device 105 is communicatively connected with another device.
[0048] Figure 3 It is a functional block diagram of the control device.
[0049] By turning on the power button, the control device 21 is started based on the power supplied from the battery 22. After starting, the control device 21 executes the control program. As a result, the control device 21 has at least an information acquisition unit 211, an adaptive control unit 212, an actuator control unit 213, a power supply unit 214, and a storage unit 215.
[0050] The information acquisition unit 211 acquires sensing information from the sole load sensor 18, the sole load sensor 19, the hip joint sensor 23, the knee joint sensor 24, the ankle joint sensor 25, and the acceleration sensor 26. The sensing information of the sole load sensor 18 and the sole load sensor 19 is load information indicating the detected load value. The sensing information of the hip joint sensor 23, the knee joint sensor 24, and the ankle joint sensor 25 is joint angle information indicating the detected joint angle. The sensing information of the acceleration sensor 26 is acceleration information indicating the detected acceleration.
[0051] The actuator control unit 213 controls the hip actuator 13 , the knee actuator 14 , and the ankle actuator 15 .
[0052] When the power button is turned on, the power supply unit 214 supplies power from the battery 22 to each part of the control device 21 .
[0053] The adaptive control unit 212 has a determination unit 2121 and a torque control unit 2122 .
[0054] The determination unit 2121 determines the repetitive motion of the user based on the acceleration detected by the acceleration sensor 26, the joint angles detected by the hip joint sensor 23, the knee joint sensor 24, and the ankle joint sensor 25, or the load value detected by the sole load sensor 18 or the sole load sensor 19. For example, the repetitive motion is a gait cycle in which one period is from the start to the end of one step.
[0055] The torque control unit 2122 compares the features extracted based on machine learning of repeated motions over multiple time periods (hereinafter referred to as multiple time periods) with a reference model for each motion. As a result, the torque control unit 2122 controls the output torque output by the hip actuator 13, the knee actuator 14, and the ankle actuator 15. Examples of features include the timing of heel strike, the timing of toe off, and the peak point of each joint torque. Heel strike is when the heel has touched the ground. Toe off refers to the situation where the toes are separated from the ground. The peak point of each joint torque is the peak time and the output value of the output torque output by the hip actuator 13, the knee actuator 14, and the ankle actuator 15, respectively.
[0056] The storage unit 215 stores a reference model corresponding to each motion having different momentums such as walking and running. The reference model is a model for determining a control mode of an output torque output by the hip actuator 13, the knee actuator 14, and the ankle actuator 15 based on a feature corresponding to each motion. The control mode is determined by parameters for calculating the output torque, etc. These parameters include, for example: control system parameters for determining the response of the hip actuator 13, the knee actuator 14, and the ankle actuator 15; and a threshold for determining standing or swinging.
[0057] For example, a reference model corresponding to walking is data stored in association with features corresponding to walking, and has parameter setting values for walking and an algorithm for walking. Also, a reference model corresponding to running is data stored in association with features corresponding to running, and has parameter setting values for running and an algorithm for running.
[0058] Note that the storage unit 215 may store reference models corresponding to other movements than walking and running. The reference model may be a reference model based on walking speed and running speed. In addition, the storage unit 215 stores angle references at each joint of each leg.
[0059] (First embodiment)
[0060] Next, the operation of the control device 21 according to the first embodiment will be described in detail.
[0061] Figure 4 is an operation block diagram showing the operation of the control device according to the first embodiment.
[0062] First, the adaptive control unit 212 controls the output torque output by the hip actuator 13, the knee actuator 14, and the ankle actuator 15 of each leg. The adaptive control unit 212 controls the output torque output by the hip actuator 13, the knee actuator 14, and the ankle actuator 15 of each leg based on the load value detected by the sole load sensor 18 and the sole load sensor 19, the acceleration Acc detected by the acceleration sensor 26, and the respective joint angles θ detected by each of the hip joint sensor 23, the knee joint sensor 24, and the ankle joint sensor 25. k , Angle reference θ k0 and the reference model stored by the storage unit 215 to control each output torque.
[0063] More specifically, first, the determination unit 2121 determines the repetitive motion of the user. A gait cycle is an example of a repetitive motion.
[0064] Figure 5 It is a diagram for explaining a method of determining a gait cycle.
[0065] The horizontal axis of the graph shown in the figure represents time. The vertical axis of the graph shown in the figure represents the magnitude of torque. The solid line 71 shown in the upper row shows an example of changes in hip joint torque. The hip joint torque is the output torque output by the hip actuator 13. The solid line 72 shown in the lower row shows an example of changes in knee joint torque. The knee joint torque is the output torque output by the knee actuator 14. In addition, the ankle actuator 15 outputs ankle joint torque. Similar to the hip joint torque and the knee joint torque, the value of the ankle joint torque also increases or decreases within a period of time according to the user's walking motion.
[0066] Furthermore, "acceleration ON" indicates that the acceleration detected by the acceleration sensor 26 is equal to or greater than a predetermined threshold. In the illustrated example, at times t1, t2, t3, t4, t5, the acceleration is equal to or greater than the predetermined threshold.
[0067] The determination unit 2121 determines the standing timing of a foot after a foot touches the ground when the acceleration exceeds a predetermined threshold value in the gait cycle from the start to the end of a step of walking. Then, the determination unit 2121 determines the standing timing of a foot to the next standing timing as a period of the gait cycle.
[0068] exist Figure 5In the example shown, the determination unit 2121 sets the duration from time t1 to time t2 as one period, time t1 being the time when the acceleration indicated by the acceleration sensor 26 becomes equal to or greater than a predetermined threshold indicating the start of a gait cycle, and time t2 being the time when the acceleration has become equal to or greater than the predetermined threshold. Similarly, the determination unit 2121 sets the duration from time t2 to time t3 as the next period, sets the duration from time t3 to time t4 as the next period, and sets the duration from time t4 to time t5 as the next period.
[0069] The determination unit 2121 outputs the gait cycle data of the most recent N (N is an integer of 2 or more) time periods to the torque control unit 2122. For example, the determination unit 2121 outputs the gait cycle data of several steps to the torque control unit 2122. The gait cycle data is the change data in the gait cycle of one time period (step). The change data shows the change of each joint angle in the gait cycle, the change of each load value, and the change of the output torque output by the hip actuator 13, the knee actuator 14, and the ankle actuator 15.
[0070] In this example, the determination unit 2121 determines the repetitive motion based on the acceleration detected by the acceleration sensor 26, but the present invention is not limited thereto. The determination unit 2121 may determine the repetitive motion based on the joint angles detected by the hip joint sensor 23, the knee joint sensor 24, and the ankle joint sensor 25. The torque control unit 2122 performs torque control processing based on the gait cycle data (N steps) of a plurality of periods.
[0071] Figure 6 is a diagram showing an overview of the torque control process.
[0072] The torque control unit 2122 inputs the gait cycle data of each step (step ST1). The torque control unit 2122 performs machine learning such as convolution processing and weighting processing on the input gait cycle data using a feature extraction model based on a convolutional neural network, and extracts features in the gait cycle (step ST2).
[0073] Examples of features include the timing of heel strike, the timing of toe off, and the peak point of joint torque. The torque control unit 2122 determines the heel strike and toe off based on the change in the load value detected by the sole load sensor 18 or the sole load sensor 19, or the change in the joint angle detected by the hip joint sensor 23, the knee joint sensor 24, and the ankle joint sensor 25. For example, when the load value exceeds a first threshold value, the torque control unit 2122 determines that the heel has touched the ground. In addition, when the load value becomes equal to or less than a second threshold value, the torque control unit 2122 determines that the toe has left. In addition, the torque control unit 2122 determines the peak point of each joint torque based on the change in the output torque output by the hip actuator 13, the knee actuator 14, and the ankle actuator 15.
[0074] Fig. 7A and Figure 7B is a graph showing an example of each feature in a gait cycle.
[0075] exist Fig. 7A and 7B In each of the graphs shown in , the horizontal axis shows the percentage of time elapsed in one gait cycle. The top row shows the change in hip joint torque. The middle row shows the change in knee joint torque. The bottom row shows the change in ankle torque. The symbol "○" indicates heel strike. The symbol "△" indicates the peak point of joint torque. The symbol "x" indicates toe-off.
[0076] Fig. 7A The changes in torque of each joint during walking are shown.
[0077] The solid line 81 shown in the upper row shows the change in the hip joint torque. The solid line 82 shown in the middle row shows the change in the knee joint torque. The solid line 83 shown in the lower row shows the change in the ankle joint torque. The start time of the unit period of the gait cycle during walking is 0% of the elapsed time in the cycle, and the end time of the unit period of the gait cycle is 100% of the elapsed time in the cycle.
[0078] Then, the duration from 0% to about 60% of the gait cycle during walking is the stance phase in which the legs are standing. The stance period refers to the period from heel strike to toe off. In addition, the duration from about 60% to 100% of the gait cycle during walking is the swing phase in which the legs are swinging. In the gait cycle during walking, the swing phase refers to the period from toe off to heel strike.
[0079] Likewise, in the gait cycle during walking, heel strike occurs at the 0% point. Additionally, in the gait cycle during walking, toe off occurs at approximately the 60% point. Additionally, the peaks of the hip joint torque and the knee joint torque in the gait cycle during walking occur at points within a period from approximately 10% to approximately 20%. The peak of the ankle joint torque in the gait cycle during walking occurs at approximately the 50% point.
[0080] Figure 7B The changes in torque at each joint during running are shown.
[0081] The solid line 91 shown in the upper row shows the change of the hip joint torque. The solid line 92 shown in the middle row shows the change of the knee joint torque. The solid line 93 shown in the lower row shows the change of the ankle joint torque.
[0082] The duration from 0% to about 35% of the gait cycle during running is the stance phase. The duration from about 35% to 100% of the gait cycle during running is the swing phase. Heel strike is at the 0% point in the gait cycle during running. Toe off is at about 35% in the gait cycle during running. In addition, in the gait cycle during running, the peak values of hip joint torque, knee joint torque, and ankle joint torque occur at points in the period from about 10% to about 20%.
[0083] like Fig. 7A and 7B As shown in Figure 2, the time ratio between the stance phase and the swing phase in the gait cycle is different between walking and running. For example, the time ratio from heel strike to toe off relative to the time of the entire gait cycle is about 60% in walking and about 40% in running. In addition, the peak time of each joint in the gait cycle is different between walking and running.
[0084] Therefore, the torque control unit 2122 compares the reference model corresponding to each motion with the extracted features and determines the user's motion ( Figure 6 For example, the torque control unit 2122 determines that the reference model having the smallest difference between the feature of the reference model and the extracted feature is the motion of the user.
[0085] Then, the torque control unit 2122 controls the output torque with the control mode according to the determined motion (step ST4). That is, the control device 21 stores a plurality of reference models in advance, specifies the reference model with the smallest difference, and obtains the control parameters of the control mode corresponding to the reference model from the storage unit 215, etc. Then, the torque control unit 2122 sets the optimal control parameters according to the difference between the determined motion and the reference model. As described above, these parameters are control system parameters for determining the response of the hip actuator 13, the knee actuator 14, or the ankle actuator 15, for determining the threshold of standing or swinging, etc.
[0086] Then, the torque control unit 2122 calculates the target value of the torque output by the hip actuator 13, the knee actuator 14, and the ankle actuator 15 based on the control parameters and the algorithm corresponding to the determined motion. Specifically, the torque control unit 2122 calculates the target value of each torque using sensing information indicating the current value obtained from at least one of the sole load sensor 18, the sole load sensor 19, the hip joint sensor 23, the knee joint sensor 24, the ankle joint sensor 25, or the acceleration sensor 26, and the angle reference of each joint. The torque control unit 2122 outputs the calculated torque target value of the hip actuator 13, the knee actuator 14, and the ankle actuator 15 of each leg to the actuator control unit 213.
[0087] The actuator control unit 213 uses the angle controller K based on the target value of the torque. ci (s) to control the rotation angles of the hip actuator 13, the knee actuator 14 and the ankle actuator 15. "s" indicates the frequency domain of the control system. Subsequently, the actuator control unit 213 controls the rotation angles of the hip actuator 13, the knee actuator 14 and the ankle actuator 15 through the force controller K bi (s) causes the hip actuator 13, knee actuator 14 and ankle actuator 15 of each leg to output torque τ at the current moment.
[0088] Thus, the interaction force F between the suit and the person applied by the user is k , the applied torque l applied by the user k , and the output torque τ in the kth (current value) of the time series becomes the power P(s) of each actuator. The hip joint sensor 23, the knee joint sensor 24, and the ankle joint sensor 25 detect each joint angle θ in the kth of the time series based on the power P(s) of the hip actuator 13, the knee actuator 14, and the ankle actuator 15 according to the power G(s) of the power suit 100. k Then, the control device 21 repeats the above-mentioned processing.
[0089] Note that the actuator control unit 213 calculates the torque τ using the following torque calculation formula as an example. In the torque calculation formula (1), "(θ) / G(s)" indicates a feedback factor to the actuator control unit 213. "F k ·l k ” indicates the feedforward factor. T i indicates a target value of the torque calculated by the torque control unit. f(θ) indicates a function including an angle θ based on an angle reference of the hip actuator 13, the knee actuator 14, and the ankle actuator 15. The torque calculation formula below is an example, and a formula other than the torque calculation formula shown below may be used. In the torque calculation formula (1), "s" represents the frequency domain of the control system, K bi represents the control model of the force controller, and K ci Indicates the control model of the angle controller.
[0090] τ=K bi (s){T i ·K ci (s)-f(θ) / G(s)+F k ·l k}(1)
[0091] Figure 8 is a flowchart illustrating the processing of the power suit according to the first embodiment.
[0092] First, the user wears the power suit 100. At this time, the user inserts the sole load sensor 19 into the foot strap 17. The sole load sensor 19 may be set in advance in the foot strap 17. As the area of the sole load sensor 19, a size suitable for the size of the user's foot may be used. In addition, the user attaches the sole load sensor 18 to the ground contact surface side of the foot strap 17 or the sole plate 16 of the power suit 100. The sole load sensor 18 may also be set in advance on the ground contact surface side of the foot strap 17 or the sole plate 16. As the area of the sole load sensor 18, a size suitable for the foot strap 17 corresponding to the size of the user's foot may be used.
[0093] The user operates a power button of the control device 21 provided in the power suit 100 to turn on the power. As a result, the control device 21 is activated. The user performs movements such as walking, running, bouncing, and jumping while wearing the power suit 100. The user can load luggage on the loading platform 20 of the power suit 100 and perform movements such as walking, running, bouncing, and jumping. The actuator control unit 213 of the control device 21 controls the hip actuator 13, the knee actuator 14, and the ankle actuator 15 to reduce the load on the user due to the weight of the luggage or the power suit 100. As a result, the power suit 100 reduces the weight of the luggage on the user and the weight of the power suit 100, and also facilitates walking.
[0094] While the control device 21 is driven, the information acquisition unit 211 acquires joint angle information from the hip joint sensor 23, the knee joint sensor 24, and the ankle joint sensor 25 at predetermined intervals (step S101). In addition, while the control device 21 is driven, the information acquisition unit 211 acquires load information from the sole load sensor 18 and the sole load sensor 19 at predetermined intervals (step S102). In addition, while the control device 21 is driven, the information acquisition unit 211 acquires acceleration information from the acceleration sensor 26 at predetermined intervals (step S103). The predetermined interval is, for example, every shorter time (such as every 10 milliseconds).
[0095] The determination unit 2121 determines the gait cycle as a repetitive motion (step S104) and generates gait cycle data based on the joint angle information, load information, and acceleration information acquired by the information acquisition unit 211. The torque control unit 2122 performs machine learning such as convolution processing and weighting processing on the gait cycle data of multiple time periods using a feature extraction model based on a convolutional neural network, and extracts features (step S105).
[0096] The torque control unit 2122 compares the extracted features with the features of the reference model (step S106). As described above, in the reference model, its features are associated with parameters and algorithms. Based on the comparison result, the torque control unit 2122 specifies a reference model in which the difference between the extracted features and the features of the reference model is smaller. The torque control unit 2122 determines the motion based on the specified reference model and sets the parameters (step S107). The torque control unit 2122 calculates the output value of the output torque using the algorithm corresponding to the determined motion (step S108).
[0097] The actuator control unit 213 causes the hip actuator 13, knee actuator 14, and ankle actuator 15 of each leg to output the torque of the calculated output value (step S109). After that, the process returns to the process of step S101, and the control device 21 repeats the process from step S101 to S109 until the process is completed.
[0098] According to the above processing, the accurate required torque of each joint for supporting the load is calculated, so that it can be output by the hip actuator 13, the knee actuator 14, and the ankle actuator 15. Therefore, even if the user's motion pattern fluctuates greatly from low response to high response (such as slow walking motion or agile running motion), the user's motion can always be tracked, and timely and appropriate assistance in reducing the load for each motion can be achieved.
[0099] For example, since the torque control unit 2122 extracts features in repetitive motions and compares the features with a reference model, the output torque can be controlled by a control mode according to the user's motion. Therefore, since the outputs of the hip actuator 13, the knee actuator 14, and the ankle actuator 15 can be controlled by adaptively predicting the user's motion, the load of the user's various motion patterns can be reduced. In addition, the torque control unit 2122 uses a feature extraction model based on a convolutional neural network, performs machine learning such as convolution processing and weighted processing, extracts features, and compares the features with a reference model. Therefore, by clarifying the difference from the benchmark, the optimal load reduction can be provided for each user.
[0100] In addition, when a device different from the power suit 100 performs the above-mentioned processing, the real-time property may be impaired due to the delay in data transmission / reception, but in the present embodiment, the control device 21 included in the power suit 100 performs the above-mentioned processing. Therefore, it is possible to respond to the user's sudden action without delay.
[0101] In addition, the above-mentioned processing is performed based on the sensing information detected by the sole load sensor 18, the sole load sensor 19, the hip joint sensor 23, the knee joint sensor 24, the ankle joint sensor 25 and the acceleration sensor 26 included in the power suit 100 attached to the lower limb. For this reason, even when there is a difference between the movements of the upper limb and the lower limb, for example, the load can be appropriately reduced. In addition, since the power suit 100 has a relatively simple structure attached to the lower limb, it can be easily attached and detached.
[0102] Although the first embodiment of the present invention has been described above, the present invention is not limited to the above embodiment, and various modifications can be made within the scope not departing from the spirit of the present invention.
[0103] For example, the sole load sensor 18 may be pre-set on one side of the ground contact surface of the sole plate 16 of the power suit 100. In addition, the sole load sensor 19 may be pre-inserted into the foot strap 17.
[0104] Furthermore, in the above description, it is shown that the sole load sensor 18 has an area covering the entire back surface of the foot band 17, and the sole load sensor 19 has an area covering the entire sole inside the foot band 17. However, the sole load sensor 18 can measure the load applied to the ground contact surface from the sole plate 16 or the foot band 17 even in the case where the position where the load is applied is deviated.
[0105] Furthermore, in the above description, a case of controlling the power suit 100 is shown, but the control device 21 is not limited thereto and may be applied to general control of a multi-joint robot or the like (eg, a humanoid robot) having nonlinear mode conversion.
[0106] Furthermore, in the above example, control device 21 is provided in power suit 100 , but the present invention is not limited thereto, and another device communicatively connected with power suit 100 in a wired or wireless manner may have the function of control device 21 .
[0107] (Second embodiment)
[0108] Fig. 9 is an operation block diagram showing the operation of the control device according to the second embodiment.
[0109] The control device 21 according to the second embodiment also detects the external torque F received from the outside by each sensor, and also uses the external torque F to control the output torque output by the hip actuator 13, the knee actuator 14, and the ankle actuator 15. The external torque F is, for example, a torque value obtained by detecting the force received from the outside by the power suit 100 by each sensor based on the impact of landing, dynamic acceleration, the movement of the power suit 100 and the deviation between the users, etc.
[0110] Here, in the second embodiment, each of the hip joint sensor 23 , the knee joint sensor 24 , and the ankle joint sensor 25 is provided with an external torque detection sensor 214 for detecting the external torque F. Then, the information acquisition unit 211 also acquires the external torque F from the external torque detection sensor 214 of each of the hip joint sensor 23 , the knee joint sensor 24 , and the ankle joint sensor 25 .
[0111] For example, the external torque detection sensor 214 detects the value of the external torque F according to the amount of change in the angle between the joint to which each joint sensor is attached and the link rotatably attached has changed due to the external force. The amount of change may be the difference between the angle indicated by the actuator control unit 213 and the angle when the change is caused instantaneously by the force obtained from the outside. The external torque detection sensor 214 may calculate the amount of change based on the amount of change in the angle between the links to which the joint rotatably attaches has changed instantaneously due to the external force, based on the difference in distances at arbitrary positions on those links instead of the difference in angles.
[0112] Fig.10 is a second flow chart illustrating processing of a power suit according to the second embodiment.
[0113] When the control device 21 is driven, the information acquisition unit 211 acquires joint angle information from the hip joint sensor 23, the knee joint sensor 24, and the ankle joint sensor 25 at predetermined intervals (step S201), and acquires the external torque F from the external torque detection sensor 214 of each sensor (step S202). When the control device 21 is driven, the information acquisition unit 211 acquires load information from the sole load sensor 18 and the sole load sensor 19 at predetermined intervals (step S203). When the control device 21 is driven, the information acquisition unit 211 acquires acceleration information from the acceleration sensor 26 at predetermined intervals (step S204). The predetermined interval is, for example, every short time (such as every 10 milliseconds).
[0114] The determination unit 2121 determines the gait cycle as a repetitive motion (step S205) and generates gait cycle data based on the joint angle information, external torque F, load information, and acceleration information acquired by the information acquisition unit 211. The torque control unit 2122 performs machine learning such as convolution processing and weighting processing on the gait cycle data of multiple cycles using a feature extraction model based on a convolutional neural network, and extracts features (step S206).
[0115] The torque control unit 2122 compares the extracted features with the features of the reference model (step S207). The torque control unit 2122 specifies a reference model having a smaller difference between the extracted features and the features of the reference model based on the comparison result. The torque control unit 2122 determines the motion and sets the parameters based on the specified reference model (step S208).
[0116] The torque control unit 2122 calculates the output value of the output torque using an algorithm corresponding to the determined motion (step S209). In calculating the output value of the output torque, the torque control unit 2122 calculates the output value of the output torque that cancels the acquired external torque F. As a result, the output value of the output torque calculated by the torque control unit 2122 becomes a value that tracks the actual torque including the dynamic element.
[0117] The actuator control unit 213 outputs the torque of the output value calculated from the hip actuator 13, the knee actuator 14, and the ankle actuator 15 of each leg (step S210). Thereafter, the process returns to the process of step S201, and the control device 21 repeats the process from steps S201 to S210 until the process is completed.
[0118] The above-described processing of the torque control unit is an example of a mode in which an output value of the torque of the drive mechanism of each sensor or the like that cancels the external torque F is calculated, and the drive mechanism is controlled based on the output value.
[0119] In the above process, the torque control unit 2122 calculates the deviation between the characteristics of the gait cycle calculated by machine learning and the characteristics of the reference model. When the deviation of the timing at which the characteristics appear is less than the threshold, the torque control unit 2122 determines the movement in step S207 and sets the parameters based on the relationship between the torque and time indicated by the reference model.
[0120] On the other hand, when the deviation of the timing at which the feature appears is equal to or greater than the threshold, the torque control unit 2122 performs the following processing. That is, the torque control unit 2122 calculates the reference model after the period change, in which the period indicated by the relationship between the torque and time of the gait cycle indicated by the reference model is offset, so that the timing of the feature of the reference model matches the gait cycle feature calculated by machine learning. Then, in step S207, the torque control unit 2122 determines the movement and sets the parameters based on the relationship between the torque and time indicated by the reference model after the period change. Thereby, the timing of the stance phase and the swing phase indicated by the reference model can be matched with the stance phase and the swing phase of the user's gait.
[0121] The processing of the torque control unit is an aspect of the processing, which corrects the time period indicated by the relationship between torque and time in the reference model based on the deviation between the features extracted based on machine learning of repetitive motion and the features of the reference model, and specifies the timing of repetitive motion according to the reference model.
[0122] If the timing of the stance phase and the swing phase indicated by the reference model can be matched with the stance phase and the swing phase of the user's walking, the period of walking can be determined without detecting the start of a step of walking based on whether the acceleration exceeds a predetermined threshold value. That is, in this case, the acceleration sensor 26 becomes unnecessary.
[0123] Fig.11 is a graph showing the relationship between the output of the torque control unit 2122 and the target torque when the control according to the second embodiment is not performed.
[0124] Fig.12 is a graph showing the relationship between the output of the torque control unit 2122 and the target torque when the control according to the second embodiment is performed.
[0125] As described above, when the external torque F is not considered, that is, the dynamically generated external torque F is not considered. To this end, the torque control unit 2122 calculates an output value based on static elements, the output value matching the torque according to the time variation indicated by the target torque p1.
[0126] On the other hand, when considering the external torque F, the dynamically generated torque is considered. Therefore, the torque control unit 2122 calculates an output value based on the dynamic element, which matches the torque that changes with time indicated by the target torque p2.
[0127] Here, when the torque control unit 2122 matches the torque corresponding to the time change indicated by the target torque p1 based on the static element, the output p3 of the torque control unit 2122 is described. Although the output p3 of the torque control unit 2122 is the torque generated at each joint due to the influence of the external torque F, under the control of the torque control unit 2122 (p3), the torque changes so as to approach the target torque p1.
[0128] However, when the output value p3 does not match the target torque p1, in the initial stage, the torque value corresponding to the time indicated by the output value p3 imposes a burden on the user of the torque difference indicated as the value pa ( Fig.11 On the other hand, the torque control unit 2122 calculates the output value p3 to match the target torque p2 ( Fig.12 ), so that the burden on users can be reduced.
[0129] Fig.13 is a diagram showing a minimum configuration of a control device.
[0130] The control device 21 as one aspect of the load reduction device may have at least the functions of the above-described determination unit 2121 and torque control unit 2122 .
[0131] The determination unit 2121 determines the repetitive motion of the user.
[0132] The torque control unit 2122 extracts features in the repetitive motions by convolution processing and compares them with the reference model of each motion. As a result, the torque control unit 2122 controls the torque output by the hip actuator 13, the knee actuator 14, and the ankle actuator 15 so as to reduce the load applied to the user at the joints of the user's legs.
[0133] The above-mentioned control device may also be a computer equipped with hardware such as a CPU (Central Processing Unit) 101 , a ROM (Read Only Memory) 102 , a RAM (Random Access Memory) 103 , a HDD (Hard Disk Drive) 104 , and a wireless communication device 105 .
[0134] The control device has a computer system inside. The process of each of the above-mentioned processes is stored in a computer-readable recording medium in the form of a program, and the process is executed by a computer that reads and executes the program. Here, the computer-readable recording medium refers to a magnetic disk, a magneto-optical disk, a CD-ROM, a DVD-ROM, or a semiconductor memory, etc. In addition, the computer program can be distributed to a computer via a communication line, and the computer that receives the distribution can execute the program.
[0135] Furthermore, the above-mentioned programs can be used to implement some of the functions described above.
[0136] Furthermore, the above-mentioned program may be a so-called differential file (differential program) that can realize the above-mentioned functions in combination with a program already recorded in the computer system.
[0137] Priority is claimed on Japanese Patent Application No. 2018-215838 filed on November 16, 2018 and Japanese Patent Application No. 2019-151162 filed on August 21, 2019, the contents of which are incorporated herein by reference.
[0138] Industrial Applicability
[0139] According to the present invention, it is possible to provide a load reducing device capable of reducing the load more appropriately with respect to the movement of the user.
[0140] [reference numerals list]
[0141] 100: Power Suit
[0142] 11: Skeleton
[0143] 12: Belt
[0144] 13: Hip Actuator
[0145] 14: Knee actuator
[0146] 15: Ankle Actuator
[0147] 16: Sole plate
[0148] 17: Foot straps
[0149] 18: Sole load sensor
[0150] 19: Foot load sensor
[0151] 20: Loading platform
[0152] 21: Control Devices
[0153] 22: Battery
[0154] 23: Hip joint sensor
[0155] 24: Knee joint sensor
[0156] 25: Ankle sensor
[0157] 26: Acceleration sensor
[0158] 211: Information acquisition unit
[0159] 212: Adaptive control unit
[0160] 214: External torque detection sensor
[0161] 2121: Determine the unit
[0162] 2122: Torque control unit
[0163] 213: Actuator control unit
[0164] 214: Power supply unit
[0165] 215: Storage unit
Claims
1. A load shedding device, include: a memory storing a plurality of reference models corresponding to each motion or according to speed or acceleration, the motion including walking or running; determining means for determining a repetitive motion repeated by a user wearing the load reduction device; and A torque control device, the torque control device is used to determining a reference model by comparing the plurality of reference models with features extracted based on machine learning using a convolutional neural network-based model for repetitive motions, the features including a timing of heel strike, a timing of toe-off, and a peak point of torque output by a drive mechanism, and Based on the determined reference mode, the torque output by the drive mechanism is controlled to reduce the load on the user at the joints of the user's legs.
2. The load reduction device according to claim 1, in, The determining means determines the start to end of one step as the repetitive motion.
3. The load reduction device according to claim 1 or claim 2, in, The determining means determines the repetitive motion based on acceleration applied to the user.
4. The load reduction device according to claim 1 or claim 2, in, The reference model relates the characteristics corresponding to each motion to a control pattern of the torque output by the drive mechanism.
5. The load reduction device according to claim 1 or claim 2, in, The torque control device calculates an output value of the torque of the drive mechanism that cancels out an external torque indicating a torque applied from the outside, and controls the drive mechanism based on the output value.
6. The load reduction device according to claim 1 or claim 2, in, The torque control device corrects the time period indicated by the relationship between torque and time in the reference model based on the following deviation, and specifies the timing of the repetitive motion based on the reference model, wherein the deviation is the deviation between the features extracted by the machine learning based on the repetitive motion and the features of the reference model.
7. A load reduction method, include: storing a plurality of reference models corresponding to each motion or according to speed or acceleration, the motion including walking or running; Determine the repetitive motions performed by users wearing load reduction equipment; as well as A reference model is determined by comparing the multiple reference models with features extracted based on machine learning using a convolutional neural network-based model for repetitive motions, the features including heel strike timing, toe-off timing, and peak points of torque output by a drive mechanism, and based on the determined reference model, the torque output by the drive mechanism is controlled to reduce the load on the user at the joints of the user's legs.
8. A storage medium storing a program for causing a computer of a load reducing device to execute a process, wherein the process include: storing a plurality of reference models corresponding to each motion or according to speed or acceleration, the motion including walking or running; determining repetitive motions repeated by a user wearing the load reduction device; as well as A reference model is determined by comparing the multiple reference models with features extracted based on machine learning using a convolutional neural network-based model for repetitive motions, the features including heel strike timing, toe-off timing, and peak points of torque output by a drive mechanism, and based on the determined reference model, the torque output by the drive mechanism is controlled to reduce the load on the user at the joints of the user's legs.
