Control method and system of gastrocnemius muscle imitating double-joint exoskeleton robot

Abstract

The invention discloses a gastrocnemius muscle imitating double-joint exoskeleton robot control method and system, and the method comprises the steps: sensing the motion state of the lower limbs of a human body based on multi-modal sensing data collection, and distinguishing different lower limb motion types through motion mode recognition. And further constructing a continuous gait phase or motion process variable to represent the current motion stage. On the basis, according to the gastrocnemius imitating double-joint action relation, cooperative power-assisted reference torque across knee joints and ankle joints is generated, and real-time power-assisted output of human body lower limb movement is achieved through an execution mechanism control module. According to the method, unified control can be achieved in various lower limb movement scenes such as periodic walking, quasi-static squatting and non-flat-ground periodic movement.

Description

Technical Field

[0001] This invention relates to the field of lower limb exoskeleton control technology, specifically to a control method and system for a gastrocnemius muscle-inspired dual-joint exoskeleton robot. Background Technology

[0002] Lower limb exoskeletons, as wearable robots that enhance human motor abilities, assist in rehabilitation training, and provide support, have broad application prospects in military, medical, industrial, and daily assistance fields. Their core control technology directly affects the safety, naturalness, and energy efficiency of human-machine collaboration.

[0003] Currently, lower limb exoskeleton control methods are mainly classified into several categories, including position control, force / torque control, hybrid control, and intention-driven control. Position control is suitable for trajectory guidance in rehabilitation training, but its rigid tracking characteristics are prone to conflict with the wearer's voluntary movements. Force control can achieve on-demand assistance, but its performance is highly dependent on high-precision force sensing and intention recognition algorithms. Hybrid control (such as impedance control) achieves a certain degree of compliance through virtual elastic-resistance models, but it still faces problems such as fixed parameters and abrupt switching in multi-joint coupling systems. Intention-driven control provides a more natural human-machine interface for exoskeletons, and can adjust the control strategy in real time according to the user's subjective intentions. However, due to the high noise and low temporal resolution of EEG signals, EEG-based exoskeleton control is not yet mature.

[0004] Furthermore, existing research largely focuses on single-joint control, typically modeling and controlling the knee and ankle joints independently. It rarely considers the cross-joint coupling characteristics and energy transfer mechanisms from the perspective of bi-joint muscle function, making it difficult to accurately reflect the biomechanical role of the gastrocnemius muscle in the coordinated regulation and energy coupling of the knee-ankle joint during gait. Simultaneously, existing control methods often employ fixed parameters or discrete mode switching impedance control, position control, or simple combinations thereof, rarely considering the continuous evolution of human-machine dynamics during gait phase and task changes. This easily leads to sudden torque changes and discontinuous human-machine interaction during phase switching or task transitions, affecting the naturalness and stability of movement. Moreover, existing control architectures often employ a structure with multiple parallel control branches but lack a unified fusion mechanism, resulting in conflicts between the outputs of different control strategies. This makes it difficult to achieve smooth and stable human-machine collaboration while ensuring the assist effect. Especially when simulating the physiological functions of the gastrocnemius muscle—such as absorbing impact during the support phase, releasing energy during the propulsion phase, and maintaining compliance during the swing phase—existing control methods often fail to achieve this continuous dynamic regulation across joints, multiple phases, and multiple tasks within a unified framework.

[0005] Therefore, there is an urgent need for a lower limb exoskeleton control method that can fully consider the characteristics of dual-joint coupling, gait phase adaptation, and smooth integration of multiple control strategies, so as to improve the biomimetic consistency, human-machine collaboration and control robustness of the exoskeleton in real-world usage scenarios. Summary of the Invention

[0006] The purpose of this invention is to provide a control method and system for a gastrocnemius muscle-inspired dual-joint exoskeleton robot, so as to at least solve the problem that traditional single-joint or discrete switching control strategies cannot simultaneously achieve both assistive effect and natural movement.

[0007] To address the aforementioned technical problems, in a first aspect, the present invention provides a control method for a gastrocnemius muscle-inspired dual-joint exoskeleton robot, comprising the following steps:

[0008] S1: Acquire multimodal sensor data to characterize the movement state of the human lower limbs;

[0009] S2: Identify the wearer's movement pattern based on the multimodal sensing data; construct gait phase variables or movement process variables to characterize the movement phase of the human lower limbs based on the movement pattern;

[0010] S3: Based on the motion pattern and its corresponding gait phase variable or motion process variable, joint torque components are generated through different control branches, and then fused through phase correlation weights to generate a single collaborative assist reference torque;

[0011] S4: The collaborative assist reference torque is input as the desired output to the exoskeleton actuator control module to perform closed-loop control and output driving torque to achieve real-time assistance for the movement of the human lower limbs.

[0012] Furthermore, the multimodal sensing data includes:

[0013] Pressure distribution data of the sole of the foot;

[0014] Exoskeleton joint angle data, angular velocity data, acceleration data, and attitude angle data;

[0015] Information on the angles and angular velocities of the knee and ankle joints.

[0016] Furthermore, motion pattern recognition methods based on rule-based models, data-driven models, or a fusion of both;

[0017] The motion patterns include:

[0018] Periodic walking movement pattern;

[0019] Quasi-static or non-periodic movement patterns, wherein the non-periodic movement patterns include standing up or squatting;

[0020] Non-flat ground periodic movement mode, which includes walking up and down stairs or ramps;

[0021] When the movement mode is a periodic walking movement mode or a non-flat ground periodic movement mode, a continuous gait phase variable is constructed, and its value range covers the complete gait cycle.

[0022] When the movement pattern is a quasi-static or non-periodic movement pattern, a movement process variable is constructed to characterize the degree of completion of lower limb movements.

[0023] Furthermore, the collaborative assistance reference torque generation method includes:

[0024] Based on the aforementioned motion pattern and its corresponding gait phase variables or motion process variables, impedance control strategy, position control strategy, and support force control strategy are used to generate the interactive torque, posture adjustment torque, and adaptive support torque of the exoskeleton joint, respectively.

[0025] Through a dynamic weighting and smooth switching mechanism, the weights of the interactive torque, attitude adjustment torque, and adaptive support torque are continuously adjusted when the gait phase changes, so that the generation results of different control strategies are weighted and fused to generate the collaborative assist reference torque.

[0026] Furthermore, the impedance control strategy includes: dynamically adjusting the virtual stiffness and virtual damping of the exoskeleton joints according to the gait phase to simulate the mechanical characteristics of the gastrocnemius muscle in the support and swing phases; and then calculating the interaction torque of the exoskeleton joints according to the virtual stiffness and virtual damping.

[0027] The position control strategy includes: calculating the angle error and angular velocity error based on the desired angle, actual angle, desired angular velocity, and actual angular velocity of the current exoskeleton joint; and then generating the attitude adjustment torque through the PD control law based on the angle error and angular velocity error.

[0028] The adaptive support force control includes: calculating the vertical support force based on the human body's center of mass acceleration and attitude angle; and then mapping the vertical support force to the adaptive support torque of the exoskeleton joint through the Jacobian matrix.

[0029] Furthermore, the dynamic weighting and smooth switching mechanism includes: assigning weight coefficients that change continuously with the task phase to different control strategies, enabling the smooth fusion of different control strategies in the time domain and phase domain, and generating the weighted and unified collaborative assist reference torque.

[0030] Furthermore, the method also includes compensating for the gravitational components generated by the exoskeleton joint itself and the load on the human body, as well as the friction of the exoskeleton joint's drive system, when generating the collaborative assist reference torque.

[0031] Furthermore, the underlying actuator includes an electric actuator, which utilizes the feedback signal from the servo drive and the linear actuator to construct a torque closed-loop PID control, enabling the actual output torque of the electric actuator to track the cooperative assist reference torque.

[0032] Secondly, the present invention provides a control system for a gastrocnemius muscle-inspired dual-joint exoskeleton robot, comprising:

[0033] The motion sensing module is used to collect multimodal sensing data;

[0034] The motion pattern recognition module is used to identify the wearer's motion pattern based on the multimodal sensing data;

[0035] The variable construction module is used to construct gait phase variables or movement process variables to characterize the movement phase of the human lower limbs based on the movement pattern.

[0036] The collaborative assistance decision module is used to generate a collaborative assistance reference torque of the exoskeleton across the knee and ankle joints based on the motion pattern and its corresponding gait phase variable or motion process variable, in accordance with the gastrocnemius muscle double joint action relationship.

[0037] The assist execution module is used to input the collaborative assist reference torque as the desired output to the exoskeleton execution mechanism control module to realize real-time assistance for the movement of the human lower limbs.

[0038] Furthermore, the motion sensing module includes:

[0039] Pressure sensors, placed on the sole of the foot or insole, are used to acquire pressure distribution data on the sole of the foot;

[0040] An inertial measurement unit (IMU) is placed on the lower limb segments or exoskeleton structure of the human body to acquire angular velocity data, acceleration data, and attitude angle data of the knee and ankle joints.

[0041] Joint angle sensors, placed at the knee and ankle joints, are used to acquire angle and angular velocity information of the knee and ankle joints.

[0042] The beneficial effects of this invention are as follows: by fusing multimodal sensing information to identify gait phase, the robustness and real-time performance of phase determination under complex gait conditions are improved; by using a multi-branch assist strategy with biomimetic impedance regulation as the core, parallel fusion of position control, adaptive support force and compensating feedforward, and phase-related dynamic weights and smooth switching mechanisms, continuous fusion of multiple control modes is achieved under a unified torque output framework to reproduce the functional characteristics of the gastrocnemius muscle in the support, energy absorption and propulsion stages; by establishing a dynamic model of the actuator and designing torque closed-loop control, high-precision tracking of the collaborative assist reference torque is achieved, thereby realizing stable collaborative control of the dual-joint exoskeleton under multi-task and multi-phase conditions. Attached Figure Description

[0043] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, use the same reference numerals to denote the same or similar parts. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0044] Figure 1 Flowchart of the control method for a gastrocnemius muscle-inspired dual-joint exoskeleton robot;

[0045] Figure 2 A schematic diagram of the control system for a gastrocnemius muscle-inspired dual-joint exoskeleton robot.

[0046] Figure 3 This is a schematic diagram of motion pattern recognition.

[0047] Figure 4 The results of the speed-torque experiment and the fitting curve of the Stribeck model are shown. Detailed Implementation

[0048] A control method for a gastrocnemius muscle (GAS)-inspired dual-joint exoskeleton robot includes the following steps:

[0049] S1: Acquire multimodal sensor data to characterize the movement state of the human lower limbs;

[0050] S2: Identify the wearer's movement pattern based on the multimodal sensing data; construct gait phase variables or movement process variables to characterize the movement phase of the human lower limbs based on the movement pattern;

[0051] S3: Based on the motion pattern and its corresponding gait phase variable or motion process variable, joint torque components are generated through different control branches, and then fused through phase correlation weights to generate a single collaborative assist reference torque;

[0052] S4: The collaborative assist reference torque is input as the desired output to the exoskeleton actuator control module to perform closed-loop control and output driving torque to achieve real-time assistance for the movement of the human lower limbs.

[0053] This method improves the robustness and real-time performance of phase determination under complex gait conditions by fusing multimodal sensing information to identify gait phase. It employs a multi-branch assist strategy centered on biomimetic impedance regulation, parallel fusion of position control, adaptive support force, and compensating feedforward. Through phase-related dynamic weights and smooth switching mechanisms, it achieves continuous fusion of multiple control modes within a unified torque output framework to reproduce the functional characteristics of the gastrocnemius muscle during support, energy absorption, and propulsion phases. By establishing a dynamic model of the actuator and designing torque closed-loop control, it achieves high-precision tracking of the collaborative assist reference torque, thereby realizing stable collaborative control of the dual-joint exoskeleton under multi-task and multi-phase conditions.

[0054] According to one embodiment of this application, the multimodal sensing data includes:

[0055] Pressure distribution data of the sole of the foot;

[0056] Exoskeleton joint angle data, angular velocity data, acceleration data, and attitude angle data;

[0057] Information on the angles and angular velocities of the knee and ankle joints.

[0058] According to one embodiment of this application, the motion pattern recognition method is based on a rule-based model, a data-driven model, or a fusion of both (such as a finite state machine method or a machine learning method); during the motion pattern recognition process, the input to the model is the aforementioned multimodal sensing data, and the output is a motion pattern label; the motion pattern includes:

[0059] Periodic walking movement pattern;

[0060] Quasi-static (stationary) or non-periodic movement patterns, wherein the non-periodic movement patterns include standing up or squatting;

[0061] Non-flat ground periodic movement mode, which includes walking up and down stairs or ramps;

[0062] When the movement mode is a periodic walking movement mode or a non-flat ground periodic movement mode, a continuous gait phase variable is constructed, and its value range covers the complete gait cycle.

[0063] When the movement pattern is a quasi-static or non-periodic movement pattern, a movement process variable is constructed to characterize the degree of completion of lower limb movements.

[0064] According to one embodiment of this application, the method for generating a collaborative assist reference torque includes:

[0065] Based on the aforementioned motion pattern and its corresponding gait phase variables or motion process variables, impedance control strategy, position control strategy, and support force control strategy are used to generate the interactive torque, posture adjustment torque, and adaptive support torque of the exoskeleton joint, respectively.

[0066] Through a dynamic weighting and smooth switching mechanism, the weights of the interactive torque, attitude adjustment torque, and adaptive support torque are continuously adjusted when the gait phase changes, so that the generation results of different control strategies are weighted and fused to generate the collaborative assist reference torque.

[0067] According to one embodiment of this application, the impedance control strategy includes: dynamically adjusting the virtual stiffness and virtual damping of the exoskeleton joint based on the gait phase to simulate the mechanical characteristics of the gastrocnemius muscle in the support and swing phases; and then calculating the interaction torque of the exoskeleton joint based on the virtual stiffness and virtual damping.

[0068] Impedance control strategy, implemented through the impedance control module, is a core component of the mid-level control strategy. Its main objective is to adjust the virtual elasticity and damping characteristics to enable the exoskeleton to exhibit similar mechanical response characteristics to the GAS during interaction with the human body. Specifically, it aims to provide high stiffness and damping during the support phase to support body weight, and reduce stiffness during the swing phase to achieve compliant following. This module replicates the GAS's function of negative energy absorption, transmission, and positive energy release at the mechanical level by constructing a mechanical mapping relationship of virtual muscles.

[0069] (1) Application of impedance control in exoskeletons

[0070] The basic idea of ​​impedance control is that the control system does not directly force position or force, but rather adjusts the "desired force-displacement" relationship to make the system exhibit specific mechanical impedance characteristics when subjected to disturbances. Its mathematical model can be described as follows:

[0071] (1);

[0072] in: To control the output force (i.e., the interaction torque) ); It is the elastic stiffness coefficient; It is the viscous damping coefficient; , which is the difference between the desired driving length and the feedback displacement.

[0073] Formula (1) defines the mechanical behavior of a "virtual tendon": when the push rod and the human body move relative to each other, the system automatically generates a restoring force to maintain balance, thereby simulating the elastic adjustment process of the GAS during the support and propulsion phases. In addition to the above-mentioned impedance control mathematical model, those skilled in the art can use other equivalent impedance expressions as needed.

[0074] (2) Phase adaptive scheduling of stiffness and damping

[0075] To reflect the time-varying characteristics of muscle dynamics during gait, impedance control parameters and It is not a constant, but depends on the gait phase. It adaptively adjusts to the motion task. Its parameter scheduling rules are defined as follows:

[0076] (2);

[0077] (3);

[0078] in: , Based on basic stiffness and damping; The force feedback coupling coefficient; This is a speed-related damping adjustment factor; This serves as a reference thrust signal. During the support phase, the system needs to exert a significant support force on the human body to counteract gravity, therefore... and The stiffness and damping remain relatively high during the oscillation phase; however, to avoid mechanical impedance interfering with the human body's movement, the stiffness and damping are significantly reduced, retaining only a small amount of compliant following. This phase-dependent impedance adjustment allows the system to automatically switch between soft and hard modes between different gait phases, thus balancing support and comfort.

[0079] (3) Force-displacement mapping and biomimicry consistency

[0080] As a biarticular muscle spanning the knee and ankle, the exoskeleton muscle (GAS) exhibits significant nonlinear force-length characteristics in its physiological function. By equating the exoskeleton pusher-linkage system to a "tendon-bone segment system," the following biomechanical mapping relationship can be established:

[0081] (4);

[0082] in: Let be the Jacobian matrix from the pushrod to the joint, reflecting the geometric transmission ratio. This equation shows that the impedance control output depends not only on the tendon length error but also on the modulation of the motion velocity, thus mechanically equivalently realizing the dynamic process of muscle contraction-relaxation.

[0083] From a biomimetic perspective, in the initial support phase, the GAS absorbs energy through eccentric contraction, exhibiting high damping and low stiffness, at which point the system effectively suppresses disturbances. During the propulsion phase, the muscles release energy through concentric contraction, corresponding to high stiffness and low damping, providing a strong output force. In the oscillation phase, the muscles relax, and stiffness and damping decrease to their minimum, allowing the system to enter a compliant following state. Therefore, the impedance regulation principle can ensure that the time-history distribution of the exoskeleton's output torque aligns with the physiological force output curve of the GAS.

[0084] (4) Dynamic smooth transition of impedance parameters

[0085] To avoid control discontinuities caused by abrupt changes in stiffness or damping parameters during phase switching, a continuously differentiable function (Sigmoid function) is used to achieve a smooth parameter transition:

[0086] (5);

[0087] in: The current gait phase (normalized to) ); The phase center; For transition rate control parameters; , This represents the stiffness value of the two-phase boundary. It can be adjusted... This allows for the control system's response speed across different phases. For example, when... When the value is large, the parameters change rapidly, which is suitable for fast gait; when When the damping is smaller, the transition is smoother, making it suitable for low-speed or rehabilitation walking scenarios. Similarly, damping... Alternatively, a smoothing function of the same form can be used. The smoothing mechanism can significantly reduce sudden torque changes and mechanical shocks, making the exoskeleton output more consistent with the human body. In addition to the Sigmoid function mentioned above, continuously differentiable functions such as the tanh function or other nonlinear transition functions can also be used.

[0088] (5) Coupling relationship between modules and the system

[0089] The impedance control module is directly connected to the upper-layer phase recognition module and the lower-layer driver closed-loop interface: it receives the task mode and phase flag values ​​from the upper layer. Downward output of collaborative assist reference torque Real-time error correction is performed through sensor feedback. In the control process, the output of the impedance module is weighted by a weighting function. Superimposed on the total torque command Together with the position control branch, the support force branch and the compensation term, they constitute the main control loop of the exoskeleton, as shown in formula (15), thereby realizing the mid-level control logic with impedance as the skeleton and task as the driver.

[0090] The impedance control module dynamically adjusts stiffness and damping between the support and swing phases to reproduce the mechanical properties of GAS biomimetic. It not only effectively improves the support capacity and compliance of the exoskeleton, but also provides a unified basic control framework for multi-mode tasks. Its adaptive scheduling and smooth transition mechanism enables the system to maintain continuity and stability in multi-phase control, and is the core component of the entire mid-level control system.

[0091] According to one embodiment of this application, the position control strategy includes: calculating the angle error and the angular velocity error based on the desired angle, the actual angle, the desired angular velocity, and the actual angular velocity of the current exoskeleton joint; and then generating the attitude adjustment torque through the PD control law based on the angle error and the angular velocity error.

[0092] The position control strategy is implemented through a position control module. The main function of this module is to provide auxiliary positioning and posture guidance capabilities to the system during gait or task phases where joint posture stability, trajectory tracking accuracy, or lower limb support balance is required. Compared to the compliant support characteristics of the impedance control module, the position control branch emphasizes geometric constraints and precise posture adjustment, making it suitable for squatting, static holding, or low-speed movement phases. In the overall control framework, this module and impedance control work synergistically in a weighted parallel manner to jointly generate a collaborative assist reference torque, achieving rigid-flexible coupling and human-machine coordination.

[0093] Physiologically, the exoskeleton system (GAS) not only has energy absorption and release functions but also participates in the stable control of the ankle-knee composite angle. In exoskeleton systems, when the human body performs movements such as squatting, standing still, or slow starting, muscles primarily work through quasi-isometric contractions. At this time, joint angle changes are relatively small, but the demand for postural stability increases significantly. The position control module targets this type of movement phase, and its control philosophy can be summarized as follows: during task phases with lower overall mechanical output, equivalent stiffness is introduced through virtual position feedback to maintain the required spatial posture and angle relationship at the joint level. This module uses a PD control law to convert angle errors and their rate of change into torque output, enabling the system to maintain a high-precision posture. The proportional-differential form of the control law is:

[0094] (6);

[0095] in: For position control torque; , The expected and actual joint angles; , The expected and actual angular velocities; , These are the proportional and differential gains, respectively.

[0096] In practical implementation, to maintain the "single torque output channel" characteristic of the mid-level control, this module does not directly drive the position servo. Instead, it treats the torque calculation result as an equivalent virtual torque source and merges it with the torque output by the impedance control module at the mid-level, as shown in formula (15). This "position error → virtual torque" approach avoids the conflict between the position closed loop and the torque closed loop, enabling the system to operate in a coordinated manner under a unified torque control framework.

[0097] From a biomechanical perspective, the GAS plays a role in posture stabilization during low-speed motion. Its control law can be viewed as a hybrid regulation of low-gain impedance and high-gain position feedback: during the static or slow squatting phase, muscles maintain the joint angle through quasi-isometric contraction, corresponding to a larger joint angle in the system. During posture adjustment or standing up, muscles undergo slight angular changes accompanied by mild concentric contraction, corresponding to higher... To suppress oscillations; during the gait swing phase, muscles relax and the exoskeleton follows compliantly, corresponding to smaller oscillations. , This physiological-control correspondence allows the exoskeleton's output to not only be dynamically consistent with the human body, but also to feel more natural.

[0098] In the system implementation, the input signals of the position control module include: the current phase. Joint angle With angular velocity The desired angle set by the upper layer Calculated Through weight A mid-level torque synthesis module is added, which, together with the impedance control output, support force, and compensation term, constitutes the final instruction.

[0099] The position control module achieves high-precision adjustment of joint angles and postures through PD-type virtual torque feedback, forming the core of the mid-level control together with the impedance control module. Through the introduction of this module, the system achieves a natural transition from force-driven dynamic support to position-driven posture maintenance, providing a solid foundation for the stability and comfort of the exoskeleton in different sports tasks.

[0100] According to one embodiment of this application, adaptive support force control includes:

[0101] Calculate the vertical support force based on the human body's center of mass acceleration and attitude angle;

[0102] The vertical support force is mapped to the adaptive support torque of the exoskeleton joint using the Jacobian matrix.

[0103] In the multi-task control system of the gastrocnemius exoskeleton, the squatting movement is one of the most challenging actions. This movement involves multiple processes, including lowering the center of gravity, energy absorption, maintaining postural balance, and releasing support force. The gastrocnemius exoskeleton (GAS) plays a complex role in dynamic support and stability control during this process. Therefore, in order to reproduce the biomimetic support characteristics of the GAS in exoskeleton control, this application designs an adaptive support force module, which is used to calculate and adjust the vertical support torque of the lower limbs in real time based on human posture and dynamic information, thereby achieving natural control of center of gravity changes.

[0104] During a squatting motion, the body's center of mass (COM) descends vertically, while the angle between the knee and ankle joints increases. The adaptive support system (GAS) primarily operates through eccentric contraction during this process: providing counter-gravity support while allowing limited angle changes for a smooth squat. Therefore, the biomimetic control strategy needs to possess the following characteristics: dynamic support adjustment capability, meaning the support force adapts to changes in posture and acceleration; energy absorption compliance, meaning it absorbs energy with low stiffness during the descent phase; posture recovery capability, meaning it provides sufficient counter-propulsion during the standing phase; and smooth transition, meaning the support force changes continuously without abrupt changes during the squat transition. Based on these requirements, the adaptive support force module calculates the vertical support force by real-time detection of the body's center of mass acceleration and posture angle information, combined with mechanical and physiological models. And converted into joint layer torque commands through geometric mapping. .

[0105] The vertical support force model of the human-exoskeleton composite system can be simplified to a mass-acceleration-attitude coupling relationship:

[0106] (7);

[0107] in: The total mass of the human body and exoskeleton; It is the acceleration due to gravity; Let be the acceleration of the center of mass in the vertical direction; The torso inclination angle is represented by this model, which reflects the dynamic changes in vertical support force with posture and motion: when... (During accelerated upward movement), the support increases; when (When squatting or descending), the supporting force decreases; when As the torso leans forward, the actual supporting force component weakens. This physical description is highly consistent with the working principle of GAS during squatting and standing: centrifugal energy absorption and centripetal energy release.

[0108] The supporting force of the exoskeleton is ultimately applied to the joints through the drive mechanism, therefore the vertical force needs to be controlled. This is converted into the corresponding joint torque. Let the instantaneous lever arm of the push rod or tendon be... The supporting moment is then calculated as follows:

[0109] (8);

[0110] in: Let be the Jacobian matrix of the push rod to joint space; This refers to the knee joint angle; The equivalent force arm varies with the angle.

[0111] In terms of control implementation, the adaptive support force module and the impedance control module operate in parallel and are superimposed on the total control torque of the middle layer through a weighted term, as shown in formula (15). This integrated structure ensures that the system can smoothly switch between walking and squatting tasks without the need for additional mode switching commands.

[0112] Table 1. Comparison of the control mechanism of the adaptive support force module with the physiological function of the gastrocnemius muscle.

[0113] Movement phase Physiological characteristics of the gastrocnemius muscle Control mechanism corresponding Squat Centrifugal contraction, energy absorption Reduce support and increase flexibility Keep Quasi-isometric contraction, stable posture Maintain stable support torque stand up Centripetal contraction provides thrust Enhanced support output to aid recovery

[0114] The control mechanism of the adaptive support force module corresponds to the physiological function of the GAS (Gastrointestinal Assay), as shown in Table 2. This biomimetic design not only replicates the energy regulation pattern of muscles at the dynamic level, but also improves comfort and the naturalness of movement at the sensory level.

[0115] According to one embodiment of this application, the dynamic weighting and smooth switching mechanism includes: assigning weight coefficients that change continuously with the task phase to different control strategies, so that different control strategies can be smoothly integrated in the time domain and phase domain, and generating the weighted and unified collaborative assist reference torque.

[0116] In the control system of the gastrocnemius exoskeleton, impedance control, position control, and adaptive support force control each perform different physiological functions: impedance control is used for compliant support and energy absorption; position control is used for posture stabilization and spatial guidance; and support force control is used for vertical load bearing and center of gravity adjustment. If a traditional mode switching method is used, the transition between different controllers often leads to sudden changes in output torque, oscillations, or discontinuous interactions, severely affecting human-machine coordination and control stability. Therefore, this application designs a dynamic weighting and smooth switching mechanism, using a continuous weighting function to achieve smooth transitions between control branches in time and phase, ensuring that the system maintains a unified torque output channel throughout the entire task domain.

[0117] The core idea of ​​the dynamic weighting mechanism is to assign weight coefficients that change continuously with the task phase to different control branches (force, position, support), so that the dominant role of the controller switches naturally in each stage, thereby achieving integrated control of "coexistence of multiple control modes - dynamic scheduling - continuous output". Mathematically, this mechanism is represented by the weighted superposition of the torques of multiple branches, as shown in formula (15). Through the continuous change of the weight function, the smooth fusion of control modes in the time domain and phase domain can be achieved, thereby ensuring the output torque. It is continuous, differentiable, and without abrupt changes throughout the entire motion cycle.

[0118] In practical implementation, the weighting function, although time... It is defined as an independent variable, but its physical meaning corresponds to gait phase or attitude angle. The evolutionary process. Due to and The two exhibit a monotonic mapping relationship within a single gait or squat cycle, and are interchangeable; therefore, this application will uniformly use the term [missing information - likely a different term]. This represents the dynamic weighting function related to its phase. To achieve dynamic scheduling of each branch control, the following weighting function is defined:

[0119] (9);

[0120] (10);

[0121] in: , The weighted transition rate parameter controls the smoothness. , The transition center phase; .

[0122] Sigmoid functions guarantee the continuity and differentiability of the weights, making them stable and controllable in numerical computation. For example, in gait cycles, The period smoothly increases from 0 (support period) to 1 (oscillation period), while Then the change is reversed; in the squatting task, As the action progresses, the force gradually increases from 0.3 to 1.0, corresponding to a shift in support force from compliant absorption to active propulsion. Based on the task characteristics, the phase switching logic of the dynamic weighting mechanism can be summarized as shown in Table 2.

[0123] Table 2 Phase switching logic of dynamic weighting mechanism

[0124] Movement phase Dominant control branch Weight configuration Control Target Support period Impedance control Smooth support and energy absorption Swing period Position control Precise trajectory and free swing squatting period Support control Energy absorption and center of gravity balance Establishment period Support + Impedance Control Active thrust and attitude recovery

[0125] This logic indicates that the system automatically adjusts the dominant control weights at different motion stages, rather than switching between discrete modes, thereby achieving the dynamic evolution of the continuous control law. This is also a key mechanism for seamless transitions in exoskeleton tasks.

[0126] The use of the Sigmoid weighting function ensures that the control output is mathematically continuous and differentiable. For each branch, the output torque... Its weighted derivative can be expressed as:

[0127] (11);

[0128] because It is a smooth function, and It is continuous within the phase, therefore the entire synthesized output Continuous differentiability throughout the entire cycle ensures that the system's dynamic response remains abrupt. Physically, this smooth switching can be viewed as a biomimetic mapping of GAS neural regulation: during the gait cycle, the excitability of motor neurons to different muscle groups is continuously distributed, and as muscles gradually transition from active contraction to passive relaxation, the system impedance changes accordingly. The smooth switching mechanism precisely reproduces this continuous regulatory process of neural control at the control level.

[0129] The dynamic weighting and smooth switching mechanism, through a continuously adjustable weighting allocation function, achieves the unified integration of three control branches: impedance, position, and adaptive support, forming a complete multi-mode control closed loop. Through this mechanism, the gastrocnemius-inspired exoskeleton achieves natural integration of multiple control modes, seamlessly switching between force control and position control, while maintaining a dynamic balance of energy and posture during vertical support force adjustment. The introduction of this mechanism endows the entire control system with continuous and coordinated characteristics similar to biological muscles, becoming a key element in achieving biomimetic compliant assistance in this system.

[0130] According to one embodiment of this application, the method further includes: compensating for the gravitational components generated by the exoskeleton joint itself and the load on the human body, as well as the friction of the exoskeleton joint's drive system, when generating the collaborative assist reference torque.

[0131] In the control system of a gastrocnemius exoskeleton, gravity and friction are the two main non-ideal factors affecting control accuracy and response stability. Gravity causes static torque bias in different postures of the exoskeleton, increasing the user's burden; while friction in the drive system (especially actuator friction and joint bearing friction) causes hysteresis and low-speed nonlinear distortion, reducing control smoothness. To address these issues, this application introduces a "compensation and feedforward module" in the mid-level control strategy to calculate and cancel these two types of disturbances in real time, thereby significantly improving the system's dynamic tracking accuracy and the naturalness of human-machine interaction.

[0132] The compensation and feedforward module is located at the bottom of the middle-level control structure. Its output is not affected by task or phase weight scheduling and is directly superimposed on the total control command in a rectangular feedforward manner, as shown in formula (15). The core idea of ​​this design is to predict and compensate for the static and dynamic non-ideal torques of the system in advance, so that the controller can use more bandwidth for the expected output of human-machine interaction, thereby achieving high-precision torque tracking and natural muscle bionic response.

[0133] (1) Gravity compensation model

[0134] In different postures, the gravitational components generated by the exoskeleton system itself and the load on the human body will be converted into joint torque offsets. If these offsets are not compensated for, it will inevitably lead to inconsistencies between the steady-state error of the control output and the auxiliary force. The basic model of gravity compensation can be written as:

[0135] (12);

[0136] in: The putter-joint Jacobian matrix; This is the mass-geometric distribution matrix; , where is the gravitational acceleration vector. The joint torque generated by gravity. This can be represented as a scalar function of joint angles. When the lower leg and foot are simplified to a point mass model, the expression is:

[0137] (13);

[0138] in: , The equivalent masses of the lower leg and foot are respectively. , This is the distance from the center of gravity to the joint. The system will [do something] within the control cycle. The GAS (Gas Assurance System) is directly added as a feedforward term to the motor commands, automatically offsetting the effects of gravity when the driver outputs its output. In the human neuromuscular system, during static posture maintenance, the GAS continuously provides baseline support force through low-frequency adjustment, consistent with the system's gravity compensation function. This compensation mechanism effectively reduces the user's active power output during static or slow movements, improving motion economy.

[0139] (2) Friction compensation model

[0140] Electric linear actuators commonly exhibit ball screw friction, gear meshing friction, and bearing friction. Their combined effect can be characterized using the Stribeck model.

[0141] (14)

[0142] in: The Coulomb friction torque; This is the static friction torque; It is the coefficient of viscous friction; Stribeck's velocity constant; For exponential shape parameters.

[0143] In the control implementation, the push rod speed is first collected. Then calculate based on the identification parameters. Finally The signal is fed forward to the control output in negative form. This means that the system automatically cancels out the friction term at the driver output stage, thus compensating for low-speed crawling and hysteresis.

[0144] This application employs a forward and reverse slow-speed scanning method: measuring the relationship between output torque and angular velocity at different speeds, and using nonlinear least-squares fitting to obtain identification parameters. The identification process is detailed below. Figure 4 Typical identification results are shown in Table 3.

[0145] Table 3 Parameter identification of the friction compensation model

[0146] parameter numerical values unit 0.0802 N·m 0.2000 N·m 1.2500 rad / s 0.0003 N·m·s / rad 3 -

[0147] Table 4. Coordination Relationships Between Control Modules

[0148] Module Input variables Output position Synergistic effect Impedance control module Displacement and velocity errors Torque Fusion Layer Provides compliance feedback Position control module Angular error Torque Fusion Layer Provides attitude adjustment ASF module Attitude and acceleration Torque Fusion Layer Provide vertical support Compensation and Feedforward Module Attitude, speed Directly superimposed on the output Counteracting non-ideal disturbances

[0149] The compensation and feedforward module, serving as the underlying support module for the mid-level control, is serially superimposed on the upper-level force-position control branch. Its main collaborative characteristics are shown in Table 4. This hierarchical torque synthesis structure ensures continuous output and smooth response across the entire motion domain, and possesses excellent anti-interference capabilities.

[0150] By introducing compensation and feedforward modules, the mid-level control system achieves high-precision torque output while ensuring biomimetic compliance, providing a solid physical foundation for the fine control of the bottom-level actuators and the overall stability of the system.

[0151] The mid-level control in this application differs from traditional single-impedance control or position control strategies. Its dual-branch parallel control structure enables adaptive shifting of the center of gravity at different motion stages: during the support phase, force control is primary, with the exoskeleton providing high mechanical impedance and support torque to assist the human body in bearing weight; during the swinging phase, position control is primary, enhancing compliance and response speed while reducing human-machine coupling stiffness; during the squatting or posture maintenance phase, the output is automatically adjusted through an adaptive support force module to achieve stable and physiologically consistent support. The overall goal of this control strategy is to achieve bidirectional, consistent human-machine mechanical interaction, meaning that the exoskeleton provides assistance without interfering with the body's natural movement patterns.

[0152] The output of the middle layer control It can be described as:

[0153] (15);

[0154] in: The interaction torque calculated for the impedance control module; The attitude adjustment torque generated for the position control branch; For squatting adaptive support torque; , These are gravity compensation and friction compensation, respectively. , , These are the weight functions.

[0155] The control law shown in formula (15) embodies the idea of ​​"single output, multi-channel fusion": the system ultimately outputs only a unified joint torque command. The control is executed by the underlying torque closed loop, while its multiple internal control branches dynamically allocate control weights through phase weighting and nonlinear scheduling. This ensures both the clarity and feasibility of the system structure, while avoiding the conflicts and interference found in traditional multi-loop control.

[0156] The gait system (GAS) plays multiple roles in the gait cycle, including absorbing impact, supporting body weight, transferring energy, and propelling the body. In a normal gait, the muscles absorb impact in the early stance phase with eccentric contraction, exhibiting high damping and low stiffness. From the stance to the propulsion phase, the muscles contract concentrically to propel the body forward, and the stiffness gradually increases. During the swing phase, the muscles relax and become more flexible to ensure the forward swing of the lower leg.

[0157] The mid-level control strategy is constructed based on this biological mechanism: the impedance control branch corresponds to the "elastic absorption-energy release" characteristics of the GAS; the position control branch corresponds to muscle relaxation and limb guidance functions; and the weight switching mechanism corresponds to the continuous transition process of muscles from active contraction to passive relaxation.

[0158] This biomimetic mapping not only makes the control laws physically reasonable, but also provides a physiological reference basis for parameter tuning, enabling the exoskeleton to have natural response characteristics in dynamic movement.

[0159] Based on the above design, the main objectives of the mid-level control strategy include:

[0160] ①Human-machine collaborative dynamic matching: Through impedance adjustment and weighted smoothing, the consistency between the exoskeleton output and the human body's torque requirements is achieved;

[0161] ② Multi-mode task adaptability: Automatically adjusts the control mode during tasks such as walking, squatting, and maintaining balance;

[0162] ③ Control the continuity and stability of the output: Achieve a continuous transition of control weights and avoid abrupt changes;

[0163] ④ Uniqueness of torque control channel: All branch control quantities are merged into a unified output in the middle layer to ensure the stability of the drive loop;

[0164] ⑤ Biomimetic consistency: The changes in each control parameter with the phase conform to the physiological characteristics of GAS, making the motion feel more natural.

[0165] According to one embodiment of this application, the underlying actuator includes an electric linear actuator, which utilizes the feedback signal from the servo drive and the linear actuator to construct a torque closed-loop PID control, enabling the actual output torque of the electric linear actuator to track and assist the reference torque.

[0166] The mid-level control strategy obtains the current gait phase and the desired joint torque under the task objective through multi-branch fusion. This torque, which integrates impedance regulation, position correction, adaptive support force, and gravity-friction compensation, is the target physical quantity that the system aims to output realistically in joint space. The task of the underlying control strategy is to ensure that the actual joint torque... The desired torque is tracked quickly, smoothly, and with high precision to complete the entire torque control closed loop. To achieve this, an execution chain of "desired torque - command torque - push rod thrust - actual joint torque" is established at the bottom layer. High-speed feedback forms a closed-loop control, ensuring the accurate implementation of the mid-level strategy at the physical drive layer. The entire bottom-level control structure can be summarized as follows:

[0167] (16);

[0168] The following explanation will focus on two aspects: dynamic modeling and closed-loop control design.

[0169] (1) Dynamic modeling of actuators

[0170] The gastrocnemius muscle-inspired exoskeleton uses a servo motor-driven electric actuator as its energy output unit. The actuator contains a motor, a lead screw-nut pair, a transmission link, and a sensing device; its dynamics can be described as follows:

[0171] (17);

[0172] in, This is the equivalent rotational inertia of the motor. This is the viscous damping coefficient of the motor. For the friction torque of the transmission system, To output electromagnetic torque to the motor, For load torque, Let be the equivalent radius of the lead screw. The linear velocity of the push rod and the angular velocity of the motor satisfy the geometric relationship of the lead screw:

[0173] (18);

[0174] in, The lead screw pitch. The push rod thrust. The mapping between the exoskeleton knee-ankle joint, applied through geometric mapping, and the actual output torque of the joint is as follows:

[0175] (19);

[0176] In the formula, The instantaneous lever arm varies with the joint angle. Let be the Jacobian matrix of the push rod-joint. Equations (17)-(19) together constitute the physical dynamics basis of the underlying actuator, reflecting the causal chain between motor speed, push rod displacement, thrust and joint torque.

[0177] (2) Torque / Thrust Closed-Loop Control Design

[0178] The goal of the underlying control is to make the electric actuator actually output torque (or thrust). Fast and smooth tracking of a given middle layer Therefore, torque closed-loop PID control is adopted. The torque error is defined. Then command torque:

[0179] (20);

[0180] The servo drive's current loop maps the command torque to the motor torque:

[0181] (twenty one);

[0182] in, The torque constant of the motor is... This is a drive current command. Considering the maximum current and power constraints of the actuator, the controller incorporates output limiting and integral anti-saturation mechanisms during execution.

[0183] (twenty two)

[0184] Integral terms are only available in Update in real time to prevent saturation accumulation from causing control lag.

[0185] Secondly, this invention provides a control system for a gastrocnemius muscle-inspired dual-joint exoskeleton robot, such as... Figure 2 As shown, it includes:

[0186] The motion sensing module is used to collect multimodal sensing data;

[0187] The motion pattern recognition module is used to identify the wearer's motion pattern based on the multimodal sensing data;

[0188] The variable construction module is used to construct gait phase variables or movement process variables to characterize the movement phase of the human lower limbs based on the movement pattern.

[0189] The collaborative assistance decision module is used to generate a collaborative assistance reference torque of the exoskeleton across the knee and ankle joints based on the motion pattern and its corresponding gait phase variable or motion process variable, in accordance with the gastrocnemius muscle double joint action relationship.

[0190] The assist execution module is used to input the collaborative assist reference torque as the desired output to the exoskeleton execution mechanism control module to realize real-time assistance for the movement of the human lower limbs.

[0191] In the hierarchical control architecture of the gastrocnemius muscle-inspired dual-joint exoskeleton, the modules form a closed-loop mechanism for coordinated operation through multiple types of signals and energy pathways. Each module is responsible for functions such as data acquisition, task identification, generation of collaborative assist reference torque, and execution tracking. The command flow, feedback flow, and energy flow among them together constitute a complete control link. Table 2 summarizes the main signals and energy paths, including the control command flow, feedback signal flow, and energy transfer link between the actuator, mechanical structure, and human body.

[0192] Table 2. Main signal and energy paths in the control system

[0193] type Start → End Key variables Function Description Control command flow Upper layer → Middle layer → Bottom layer Characterizing the decision path from task intent to physical torque Feedback signal flow Bottom layer → Middle layer → Top layer Provides mechanical and motion state feedback to achieve closed-loop correction. Energy Flow Drivers ↔ Exoskeleton Mechanical Structure ↔ Human Musculoskeletal System Electrical work → Mechanical work (thrust) → Joint torque → Muscle force response Reflecting the dynamics of power exchange and coupling between humans and machines

[0194] The control command flow runs through a three-layer structure from top to bottom: the upper layer identifies gait phase and task mode based on multimodal sensor information; the middle layer generates a unified desired torque through modules such as impedance adjustment, position adjustment, adaptive support force, and compensation feedforward; and the lower layer executes thrust and current control based on this desired torque to achieve physical output. The feedback signal flow transmits state information such as thrust, joint angle, and angular velocity from the lower layer upwards, providing necessary error correction for the middle and upper layers, ensuring the stability and adaptability of the control process. The energy flow reflects the power distribution of the system during thrust execution, joint torque output, and human response. Through the coupling between the actuator, mechanical structure, and musculoskeletal system, the biomimetic characteristics of GAS (Gas Assault System) negative energy absorption and positive energy release are achieved.

[0195] According to one embodiment of this application, the system further includes a data acquisition module, which includes pressure sensors, an inertial measurement unit (IMU), and joint angle sensors (encoders) distributed on the sole of the foot. The IMU unit is arranged on the lower limb segments or exoskeleton structure of the human body to acquire angular velocity, linear acceleration, and attitude information of the corresponding parts; the pressure sensors are arranged on the sole of the foot or insole to acquire pressure or load signals on the sole of the foot or insole to characterize the foot contact state and support distribution;

[0196] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A control method for a gastrocnemius muscle-inspired dual-joint exoskeleton robot, characterized in that, Includes the following steps: S1: Acquire multimodal sensor data to characterize the movement state of the human lower limbs; S2: Identify the wearer's movement pattern based on the multimodal sensing data; construct gait phase variables or movement process variables to characterize the movement phase of the human lower limbs based on the movement pattern; S3: Based on the motion pattern and its corresponding gait phase variable or motion process variable, joint torque components are generated through different control branches, and then fused through phase correlation weights to generate a single collaborative assist reference torque; S4: The collaborative assist reference torque is input as the desired output to the exoskeleton actuator control module to perform closed-loop control and output driving torque to achieve real-time assistance for the movement of the human lower limbs.

2. The control method for the gastrocnemius muscle-inspired dual-joint exoskeleton robot according to claim 1, characterized in that, The multimodal sensing data includes: Pressure distribution data of the sole of the foot; Exoskeleton joint angle data, angular velocity data, acceleration data, and attitude angle data; Information on the angles and angular velocities of the knee and ankle joints.

3. The control method for the gastrocnemius muscle-inspired dual-joint exoskeleton robot according to claim 2, characterized in that, The motion pattern recognition method includes a motion pattern recognition method based on a rule-based model, a data-driven model, or a fusion of both. The motion patterns include: Periodic walking movement pattern; Quasi-static or non-periodic movement patterns, wherein the non-periodic movement patterns include standing up or squatting; Non-flat ground periodic movement mode, which includes walking up and down stairs or ramps; When the movement mode is a periodic walking movement mode or a non-flat ground periodic movement mode, a continuous gait phase variable is constructed, and its value range covers the complete gait cycle. When the movement pattern is a quasi-static or non-periodic movement pattern, a movement process variable is constructed to characterize the degree of completion of lower limb movements.

4. The control method for the gastrocnemius muscle-inspired dual-joint exoskeleton robot according to claim 3, characterized in that, The method for generating the collaborative assist reference torque includes: Based on the aforementioned motion pattern and its corresponding gait phase variables or motion process variables, impedance control strategy, position control strategy, and support force control strategy are used to generate the interactive torque, posture adjustment torque, and adaptive support torque of the exoskeleton joint, respectively. Through a dynamic weighting and smooth switching mechanism, the weights of the interactive torque, attitude adjustment torque, and adaptive support torque are continuously adjusted when the gait phase changes, so that the generation results of different control strategies are weighted and fused to generate the collaborative assist reference torque.

5. The control method for the gastrocnemius muscle-inspired dual-joint exoskeleton robot according to claim 4, characterized in that, The impedance control strategy includes: dynamically adjusting the virtual stiffness and virtual damping of the exoskeleton joints according to the gait phase to simulate the mechanical characteristics of the gastrocnemius muscle in the support and swing phases; and then calculating the interaction torque of the exoskeleton joints according to the virtual stiffness and virtual damping. The position control strategy includes: calculating the angle error and angular velocity error based on the desired angle, actual angle, desired angular velocity, and actual angular velocity of the current exoskeleton joint; and then generating the attitude adjustment torque through the PD control law based on the angle error and angular velocity error. The adaptive support force control includes: calculating the vertical support force based on the human body's center of mass acceleration and attitude angle; and then mapping the vertical support force to the adaptive support torque of the exoskeleton joint through the Jacobian matrix.

6. The control method for the gastrocnemius muscle-inspired dual-joint exoskeleton robot according to claim 4 or 5, characterized in that, The dynamic weighting and smooth switching mechanism includes: assigning weight coefficients that change continuously with the task phase to different control strategies, so that different control strategies can be smoothly integrated in the time domain and phase domain, and generating the weighted and unified collaborative assist reference torque.

7. The control method for the gastrocnemius muscle-inspired dual-joint exoskeleton robot according to claim 1, characterized in that, The method also includes compensating for the gravitational components generated by the exoskeleton joint itself and the load on the human body, as well as the friction of the exoskeleton joint's drive system, when generating the collaborative assist reference torque.

8. The control method for the gastrocnemius muscle-inspired dual-joint exoskeleton robot according to claim 1, characterized in that, The underlying actuator includes an electric actuator, which uses the feedback signal from the servo drive and the linear actuator to construct a torque closed-loop PID control, so that the actual output torque of the electric actuator tracks the cooperative assist reference torque.

9. A control system for a gastrocnemius muscle-inspired dual-joint exoskeleton robot, characterized in that, include: The motion sensing module is used to collect multimodal sensing data; The motion pattern recognition module is used to identify the wearer's motion pattern based on the multimodal sensing data; The variable construction module is used to construct gait phase variables or movement process variables to characterize the movement phase of the human lower limbs based on the movement pattern. The collaborative assistance decision module is used to generate a collaborative assistance reference torque of the exoskeleton across the knee and ankle joints based on the motion pattern and its corresponding gait phase variable or motion process variable, in accordance with the gastrocnemius muscle double joint action relationship. The assist execution module is used to input the collaborative assist reference torque as the desired output to the exoskeleton execution mechanism control module to realize real-time assistance for the movement of the human lower limbs.

10. The control system for the gastrocnemius muscle-inspired dual-joint exoskeleton robot according to claim 9, characterized in that, The motion sensing module includes: Pressure sensors, placed on the sole of the foot or insole, are used to acquire pressure distribution data on the sole of the foot; An inertial measurement unit (IMU) is placed on the lower limb segments or exoskeleton structure of the human body to acquire angular velocity data, acceleration data, and attitude angle data of the knee and ankle joints. Joint angle sensors are placed at the knee and ankle joints to acquire angle and angular velocity information of the knee and ankle joints.

Images