Lower limb exoskeleton man-machine synchronization adaptive admittance control method based on disturbance observer

Abstract

The invention discloses a lower limb exoskeleton man-machine synchronization self-adaptive admittance control method based on a disturbance observer, and provides a torque-sensor-free control scheme for solving the problems that an existing lower limb exoskeleton is large in intention recognition delay, high in model dependence and prone to drifting of a physical sensor. The method comprises the following steps: firstly, establishing a man-machine coupling dynamic model of a single-joint equivalent rigid body, and uniformly representing human body active torque, nonlinear friction and external load change as total disturbance of a system; secondly, constructing a three-order linear extended state observer, observing the total disturbance in real time by using a joint encoder and a control instruction, and reconstructing a man-machine interaction torque through a model-assisted disturbance separation strategy; furthermore, a variable parameter admittance controller is constructed in combination with a gait phase calculated by an inertial measurement unit, and the virtual stiffness and the damping coefficient are dynamically adjusted according to the observed interaction torque amplitude. According to the invention, accurate perception and compliant response to human body motion intentions can be realized on a low-computing-power embedded platform, the phenomenon of asynchronous man-machine motion is effectively eliminated, and the comfort and walking metabolism efficiency of a wearer are remarkably improved while the robustness of the system is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of robot control, in particular to a lower limb exoskeleton human-robot synchronous adaptive admittance control method based on a disturbance observer. BACKGROUND

[0002] As a typical human-robot coupling system, lower limb exoskeleton robots are widely used in medical rehabilitation, individual combat and industrial transportation fields. The core technology is how to realize the accurate identification and compliant following of the wearer's motion intention by the robot, that is, to realize the synchronous motion of "human-robot integration". However, the existing lower limb exoskeleton human-robot interaction control technology still faces many technical bottlenecks in practical application, mainly in three aspects of hardware dependency, model uncertainty and environmental adaptability.

[0003] Firstly, in the hardware implementation of motion intention recognition, the existing technology excessively relies on expensive physical sensors. Most active exoskeletons usually install multi-dimensional force / torque sensors at the joints or install pressure sensor arrays on the soles to calculate the human motion intention by measuring the human-robot contact force. However, such high-precision sensors not only have high cost and fragile structure, but also have high installation precision requirements. More seriously, physical torque sensors generally have temperature drift and zero drift problems, and are easily disturbed by noise in complex electromagnetic environments. In order to suppress noise, the control system usually needs to introduce a low-pass filter, which inevitably brings signal phase lag. This "perception delay" directly leads to the response of the robot always being slower than the human action half beat, making the wearer feel obvious "dragging" and heavy physical burden, and it is difficult to realize real-time synchronization at the microsecond level.

[0004] Secondly, in the theoretical basis of the control algorithm, the existing method has too strong dependence on the accurate physical model. Traditional model-based control methods (such as computed torque method, inverse dynamics control) need to establish an extremely accurate human-robot coupling dynamics model. However, in actual application scenarios, the human body is a time-varying complex biomechanical system, different wearers have different heights, weights and limb inertias, and the same wearer's dynamics parameters will also change significantly under the conditions of weight change (such as carrying heavy objects) or muscle fatigue. Traditional algorithms are difficult to obtain these time-varying physical parameters in real time, and once the preset model parameters deviate from the actual system, the compensation performance of the controller will decrease sharply, leading to system oscillation or instability, which seriously affects the robustness and safety of the system.

[0005] Finally, regarding the adaptability of the interaction strategy, traditional admittance / impedance control often uses fixed parameters, lacking the ability to adapt to gait and terrain. Existing admittance control methods typically set the stiffness and damping coefficient of human-machine interaction to constants. However, human lower limb walking is a periodic process of varying stiffness: high stiffness is required in the support phase to provide stable support, while low damping and low stiffness are required in the swing phase to achieve light and agile strides. Fixed-parameter controllers cannot meet these two contradictory needs: if the parameters are set too stiff, the swing leg will experience huge motion resistance; if the parameters are set too soft, the support phase will experience a "softening" phenomenon. In addition, when encountering external disturbances caused by complex terrain (such as slopes, uneven surfaces), fixed parameters cannot dynamically adjust compliance according to the magnitude of the disturbance, causing the robot to be unable to make adaptive adjustments like a human's instinctive reaction, reducing the metabolic efficiency and comfort of walking.

[0006] In summary, developing a lower limb exoskeleton control method that no longer relies on expensive and drift-prone physical torque sensors, can break free from dependence on precise human dynamics models, and can adaptively adjust control parameters in real time based on gait phase and external disturbances is a key technical problem that urgently needs to be solved in this field. Summary of the Invention

[0007] To address the aforementioned problems, this invention proposes a human-machine synchronous adaptive admittance control method for lower limb exoskeleton based on a disturbance observer. The specific steps are as follows, characterized by:

[0008] Step 1: Optimize the exoskeleton hardware architecture and embedded computing platform, using the TMS320F28377D as the core control unit, and using joint encoders and inertial measurement units to collect motion status.

[0009] Step 2: Construct a human-machine coupled dynamics model and disturbance definition framework, and establish a mathematical model in the control system that includes the dynamics of the exoskeleton rigid body and the motion characteristics of the human body; define the torque actively applied by the human body, joint friction, ground reaction force and complex factors of sudden external load changes as the total disturbance of the system.

[0010] Step 3: Deploy a linear adaptive disturbance observer and an extended state observer. Use the extended state observer to observe and estimate the total disturbance defined in Step 2 in real time. Through rapid iteration of the internal state of the algorithm, calculate the current human-computer interaction torque value, and realize the function of replacing the hardware torque sensor with a software algorithm.

[0011] Step 4: Construct a variable-parameter admittance controller based on gait phase, solve the wearer's gait cycle in real time using inertial measurement unit and encoder data; design a dynamic mapping rule to adaptively adjust the virtual stiffness and damping coefficients of the admittance controller in real time according to the disturbance value observed by the extended state observer and the current gait phase, so as to ensure that the robot can conform to the motion intention of the human body under complex terrain, and realize the compliant interaction of walking when the human walks and stopping when the human stops.

[0012] The lower limb exoskeleton human-machine synchronous adaptive admittance control method based on a disturbance observer has the beneficial effects and technical effects that:

[0013] 1. The software algorithm of the linear extended state observer (ESO) realizes the "soft measurement" of the human-machine interaction torque, so as to replace the expensive, fragile and complex installation and maintenance of the physical multi-dimensional force / torque sensor. This not only significantly reduces the cost of the exoskeleton system, but also fundamentally eliminates the problems of zero drift, temperature drift and signal noise of the physical sensor, greatly improving the long-term running stability of the system in harsh working conditions.

[0014] 2. The present application solves the problem of model uncertainty caused by the large individual difference of human body parameters and the change of load (such as carrying heavy objects), by defining the internal model dynamics, friction nonlinearity and external load change as total disturbance, and using LADRC technology for real-time observation and compensation. This method makes the control system not need to obtain accurate human body inertia or damping parameters in advance, and can realize high-performance control, solving the technical problem that the traditional model-based control algorithm is sensitive to parameter perturbation and easy to lose stability.

[0015] 3. The variable-parameter admittance controller based on gait phase and disturbance amplitude can simulate the variable stiffness characteristics of human biological limbs: providing high stiffness stable support in the support phase, providing low damping compliant following in the swing phase, and automatically reducing the stickiness when strong acceleration intention is perceived. This dynamic adjustment mechanism ensures that the robot not only walks smoothly on flat ground, but also automatically and flexibly adapts to uphill and downhill or road impact. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The flowchart of the present application;

[0017] Figure 2 The ESO algorithm flowchart of the present application. DETAILED DESCRIPTION

[0018] The present application will be further described in detail below in combination with the drawings and specific embodiments:

[0019] The application provides a lower limb exoskeleton human-machine synchronous adaptive admittance control method based on a disturbance observer. Figure 1 The steps of the application are described in detail below.

[0020] Step 1: Optimize the exoskeleton hardware architecture and embedded computing platform, use TMS320F28377D as the core control unit, and use joint encoders and inertial measurement units to collect motion states.

[0021] Step 2: Construct a human-machine coupled dynamics model and a disturbance definition framework, establish a mathematical model containing exoskeleton rigid body dynamics and human motion characteristics in the control system; define the torque, joint friction, ground reaction force and external load sudden change complex factors applied by the human body as the total disturbance of the system.

[0022] Step 2 establishes a mathematical relationship describing the cooperative motion of the lower limb exoskeleton and the human body, and all model uncertainties and external forces are collectively processed, and the specific operation is as follows:

[0023] Step 2.1 establishes the Lagrange dynamics equation of the human-machine system

[0024] Based on the hardware architecture described in step 1, a single joint is regarded as a rigid link system. The complex flexibility is ignored, and according to the Lagrange dynamics principle, the following lower limb exoskeleton human-machine coupled nonlinear dynamics equation is established:

[0025]

[0026] wherein, , , , and represent the angle, angular velocity and angular acceleration of the joint, respectively. is the equivalent inertia matrix containing the exoskeleton and the human limb; is the Coriolis force and centrifugal force term; is the gravity torque term; is the Coulomb friction and viscous friction torque at the joint; is the control torque applied by the motor; is the interactive torque actively applied by the human body;

[0027] Step 2.2 model decomposition and parameter reconstruction

[0028] To get rid of the dependence on the precise physical model parameters, the above equation is rewritten in the standard form of a second order system. The control input gain part is separated, and all the remaining nonlinear terms, coupling terms and external torques are lumped into an unknown term.

[0029] Multiply both sides of the equation by , and rearrange to get

[0030]

[0031] Introduce a control gain parameter , which is an approximation of the system control input gain. Define the deviation of the actual system from the approximate model as the total disturbance. The state equation of the system can now be simplified as

[0032]

[0033] In this equation, represents the control input, i.e. , is the total disturbance of the system defined in this invention. Its mathematical definition is as follows:

[0034]

[0035] Step 2.3 Establish an augmented state space model for observer design

[0036] To observe the total disturbance in step 3 using a linear adaptive disturbance observer, the total disturbance needs to be augmented into a new state variable. Define the angle state variable , the angular velocity , and introduce the total disturbance augmented state . Assume that the total disturbance is differentiable, and denote its derivative as .

[0037] The continuous-time state space model of the human-machine coupled system is constructed as follows:

[0038]

[0039] Through this step, the nonlinear, strongly coupled human-machine system is transformed into an integral series type linear system containing an unknown disturbance term . This mathematical transformation implicitly includes the human's active motion intention in the state , so it is not necessary to directly measure through an expensive torque sensor. Instead, the state By performing real-time estimation, all dynamic information, including human-computer interaction intentions, can be obtained, laying a theoretical foundation for subsequent sensorless admittance control.

[0040] Step 3: Deploy a linear adaptive disturbance observer and an extended state observer. Use the extended state observer to observe and estimate the total disturbance defined in Step 2 in real time. Through rapid iteration of the internal state of the algorithm, calculate the current human-computer interaction torque value, realize the function of replacing the hardware torque sensor with a software algorithm, and solve the problems of large delay and sensor signal drift in traditional intent recognition.

[0041] This step aims to design and implement a linear extended state observer in an embedded controller based on the extended state-space model constructed in step 2. The core of this step lies in reconstructing the system state and total disturbance in real time using an algorithm. The flowchart of the extended state observer is shown below. Figure 2 As shown, the specific operation is as follows:

[0042] Step 3.1 Constructing the mathematical model of the linear extended state observer

[0043] Based on the continuous-time state equation derived in step 2, a full-dimensional state observer is constructed. This is achieved using real-time angle signals acquired by the joint encoder. As input to the observer, define , , From the system perspective angular velocity and total disturbance The estimated value.

[0044] Design an observer equation of the following form:

[0045]

[0046] in, The joint angle is measured by the sensor. This is the current motor control input; This refers to the system state observation error; , , The observer gain coefficient determines the observer's speed of disturbance tracking and its noise suppression capability. , , They are respectively , , The derivative of .

[0047] Through this equation, the system explicitly transforms the total disturbance into state variables. It performs direct calculations without resorting to inverse physical models.

[0048] Step 3.2 Observer parameter tuning based on bandwidth method and discretization implementation

[0049] In order to ensure the engineering realizability and stability of the algorithm on the embedded platform, the bandwidth parameterization method is used to configure the observer gain. The characteristic equation of the observer is configured to have the same pole , so that the three undetermined parameters are simplified to a unique adjustment parameter .

[0050] The gain calculation formula is as follows:

[0051]

[0052] In the TMS320F28377D dual-core DSP, the zero-order hold method is used to discretize the above continuous equation. The discrete iterative operation is performed in the interrupt service subroutine, which ensures that the observer converges much faster than the system dynamic response speed, so as to realize the non-delay estimation of the total disturbance .

[0053] Step 3.3 Interaction torque analysis

[0054] The disturbance estimation value output by the observer is converted into human-machine interaction torque with clear physical meaning , replacing the expensive hardware torque sensor. According to the definition of step 2, the total disturbance is mainly composed of internal model dynamics and external torque. In the case of low-speed operation of the system or known inertia parameters, the human-machine interaction torque is separated from and solved by the following formula:

[0055]

[0056] where is the nominal inertia estimation value; is the human-machine interaction torque; is the gravity compensation term.

[0057] Step 4: Construct a variable-parameter admittance controller based on gait phase, use inertial measurement unit and encoder data to solve the wearer's gait cycle in real time; design a dynamic mapping rule to adaptively adjust the virtual stiffness and damping coefficients of the admittance controller in real time according to the disturbance value observed by the extended state observer and the current gait phase, ensuring that the robot can conform to the human's motion intention in complex terrain, realizing the compliant interaction of walking when the human walks and stopping when the human stops.

[0058] This step aims to establish the human-robot interaction logic layer of the exoskeleton, and convert the generalized disturbance observed in step 3 into robot motion instructions. By simulating a parameter-variable virtual spring-damping system, the robot can automatically adjust the softness and hardness according to the walking intention and gait phase of the human, and the specific operation is as follows:

[0059] Step 4.1 Real-time gait phase solution and state mechanism construction

[0060] Collect the Euler angles of the thigh and lower leg using the inertial measurement unit configured in step 1 , Collect joint angles using joint encoders , Construct a finite state machine to divide the gait cycle.

[0061] Set threshold decision logic:

[0062] Support phase: when the foot is in contact with the ground and the joint angular velocity is approximately zero or within a certain support interval, it is determined that high impedance is required.

[0063] Swing phase: when the limb moves in the forward direction and the acceleration is obvious, it is determined that low impedance is required.

[0064] Introduce gait phase variable , where 0 represents complete swing phase and 1 represents complete support phase, and design a smooth transition Sigmoid function between the two to avoid system oscillation caused by sudden changes in control parameters.

[0065] Step 4.2 Construct a virtual admittance control model

[0066] Establish the desired dynamic relationship equation between the exoskeleton end effector and the wearer. This method takes the total disturbance estimate output by the extended state observer in step 3 as the input of the admittance model, is the scaled interaction torque after the inertia matrix . .

[0067] The admittance control law is defined as follows:

[0068]

[0069] where, is the reference expected trajectory calculated by the admittance controller that the exoskeleton should follow; is the equilibrium position under the action of no external force; is the virtual inertia parameter; , are the time-varying virtual damping coefficient and virtual stiffness coefficient, respectively; is the generalized disturbance observed in step 3 solved real-time human-robot interaction torque estimate; , the first derivative and the second derivative of respectively.

[0070] Step 4.3 Designing variable parameter mapping rule based on disturbance observation and gait phase

[0071] In order to achieve the adaptive effect of walking when the person walks and stopping when the person stops, the dynamic adjustment strategy of and is designed. This strategy depends on the gait phase and the size of the disturbance observation value :

[0072] Designing stiffness adjustment strategy :

[0073]

[0074] wherein, is the minimum stiffness threshold; is the maximum stiffness threshold. In the support phase increase the stiffness to provide load support; in the swing phase reduce the stiffness to reduce the motion resistance.

[0075] Designing damping adjustment strategy :

[0076] Use the amplitude of the disturbance observation value to judge the strength of the person's motion intention.

[0077]

[0078] wherein, is the reference damping; is the adjustment factor. When the observer detects a larger interaction torque, the algorithm automatically reduces the virtual damping , reduces the stickiness of the system, thereby speeding up the response speed of the exoskeleton to human action, and eliminating the out-of-sync feeling between man and machine.

[0079] Step 4.4 Closed-loop control command generation and execution

[0080] The expected position and the expected velocity calculated by the admittance model are used as the tracking target of the underlying linear adaptive disturbance observer controller, that is, the command basis of in step 2.

[0081] The final execution logic of the control system is:

[0082] Step 4.4.1 The extended state observer outputs the disturbance in real time ;

[0083] Step 4.4.2 The admittance controller calculates the reference trajectory according to and gait phase, which complies with human intention ;

[0084] Step 4.4.3 The input position loop drives the motor to accurately track the trajectory again by using the high anti-interference characteristics of the linear adaptive disturbance observer itself, while compensating for external friction and gravity.

[0085] Through this step, the system not only has anti-interference ability at the bottom layer, but also has intelligent characteristics of understanding human intention and actively cooperating at the upper layer, which can solve the problem of traditional control being rigid and dragging under complex terrain.

[0086] The above description is only a preferred embodiment of the present application, and does not limit the present application in any other form, and any modification or equivalent change made according to the technical essence of the present application still falls within the scope of the present application.​

Claims

1. A lower extremity exoskeleton human-robot synchronization adaptive admittance control method based on a disturbance observer, the specific steps being as follows, characterized in that: Step 1: Optimize the exoskeleton hardware architecture and embedded computing platform, use TMS320F28377D as the core control unit, and use joint encoders and inertial measurement units to collect motion states; Step 2: Build a human-robot coupling dynamics model and a disturbance definition framework, establish a mathematical model in the control system that includes the rigid body dynamics of the exoskeleton and the motion characteristics of the human body; define the torque, joint friction, ground reaction force and external load mutation complex factors applied by the human body as the total disturbance of the system; Step 3: Deploy a linear adaptive disturbance observer and an extended state observer, which is used to observe and estimate the total disturbance defined in step 2 in real time; by rapidly iterating the internal state of the algorithm, the current human-robot interaction torque value is calculated, realizing the function of replacing the hardware torque sensor with a software algorithm; Step 4: Build a variable parameter admittance controller based on gait phase, use the inertial measurement unit and encoder data to calculate the wearer's gait cycle in real time; design a dynamic mapping rule to adaptively adjust the virtual stiffness and damping coefficients of the admittance controller according to the disturbance value observed by the extended state observer and the current gait phase, ensuring that the robot can conform to the human's motion intention in complex terrain, realizing the compliant interaction of walking when the human walks and stopping when the human stops.

2. The disturbance observer based lower extremity exoskeleton human-robot synchronization adaptive admittance control method according to claim 1, characterized in that: The mathematical relationship describing the coordinated motion of the lower extremity exoskeleton and the human body in step 2 can be expressed as: Step 2.1 Establish the Lagrange dynamics equation of the human-robot system Based on the hardware architecture described in step 1, a single joint is regarded as a rigid link system; ignoring complex flexibility, according to the Lagrange dynamics principle, the following lower extremity exoskeleton human-robot coupling nonlinear dynamics equation is established: ； wherein, , , respectively represent the angle, angular velocity and angular acceleration of the joint; is the equivalent inertia matrix of the exoskeleton and the human limb; is the Coriolis and centrifugal force term; is the gravity torque term; is the Coulomb and viscous friction torque at the joint; is the control torque applied by the motor; is the interaction torque actively applied by the human body; is the unknown external environmental disturbance; Step 2.2 Model decomposition and parameter reconstruction In order to get rid of the dependence on accurate physical model parameters, the above equation is rewritten into the standard form of a second-order system; the control input gain part is separated, and all other nonlinear terms, coupled terms and external torques are unified into unknown terms; Multiplying both sides of the equation by , we get ； Introducing a control gain parameter which is an approximation of the system control input gain; the deviation of the actual system from the approximate model is defined as the total disturbance; at this point, the state equation of the system can be simplified as ； In this equation, represents the control input, i.e. , total disturbance defined for the system of the present invention; its mathematical definition is as follows: ； Step 2.3 Establish an extended state space model for observer design In order to utilize the linear adaptive disturbance observer for the total disturbance in step 3 To conduct observations, it is necessary to... Expand it into a new state variable; define the angle state variable. angular velocity And introduce the total disturbance extended state. Assume the total disturbance is differentiable, and denote its derivative as... ; The continuous-time state space model of the human-robot coupling system is constructed as follows: ； This step transforms the nonlinear, strongly coupled human-machine system into one containing unknown perturbation terms. Integral-series linear systems; this mathematical transformation will enable the human body to actively move. Implicit in state This eliminates the need for direct measurement via expensive torque sensors. In step 3, simply extend the state observer to monitor the state. By performing real-time estimation, all dynamic information, including human-computer interaction intentions, can be obtained.

3. The disturbance observer based lower extremity exoskeleton human-robot synchronization adaptive admittance control method according to claim 1, characterized in that: The linear adaptive disturbance observer and the extended state observer deployed in step 3 can be expressed as follows: Step 3.1 Construct the mathematical model of the linear extended state observer Based on the continuous-time state equation derived in Step 2, construct the full-state observer; utilize the real-time angle signals collected by the joint encoders As the input of the observer, define , , are the estimated values of the system angles , angular velocities and total disturbances , respectively; The observer equation is designed as follows: ； wherein, is the joint angle measured by the sensor; is the current motor control input; is the system state observation error; , , is the observer gain coefficient, which determines the speed of the observer tracking the disturbance and the noise suppression ability; , , are the derivatives of , , respectively; With this equation, the system explicitly transforms total disturbances into state variables Direct calculation is performed without back-solving through physical models; Step 3.2 Observer parameter tuning and discretization implementation based on bandwidth method In order to ensure the engineering realizability and stability of the algorithm on the embedded platform, the bandwidth parameterization method is used to configure the observer gain; the characteristic equation of the observer is configured as the same pole , so as to simplify the three undetermined parameters into the only adjustment parameter . The gain calculation formula is as follows: ； In the TMS320F28377D dual-core DSP, the above continuous equation is discretized by using the zero-order hold method; the discrete iterative operation is performed in the interrupt service subroutine, which ensures that the convergence speed of the observer is much faster than the dynamic response speed of the system, thereby realizing the non-delay estimation of the total disturbance . Step 3.3 Interaction torque analysis The disturbance estimate from the observer output is transformed into a human-machine interaction torque with a clear physical meaning , replacing the expensive hardware torque sensor; according to the definition of step 2, the total disturbance is mainly composed of the internal dynamics of the model and the external torque; in the case of low speed of the system or known inertia parameters, the human-machine interaction torque is isolated from by the following equation: ； wherein, is a nominal inertia estimate value; is a human interaction torque; is a gravity compensation term.

4. The disturbance observer based lower extremity exoskeleton human-robot synchronization adaptive admittance control method according to claim 1, characterized in that: The variable parameter admittance controller based on gait phase constructed in step 4 is expressed as follows: Step 4.1 Real-time gait phase calculation and state construction Collecting thigh, shank Euler angles with IMU configured in step 1 , , Collecting joint angles with joint encoders , , Building finite state machine to partition gait cycle; Set the threshold decision logic: Support phase: when the foot bottom contacts the ground and the joint angular velocity is approximately zero or in a specific support interval, it is determined as a high impedance demand state; Swing phase: when the limb moves in the forward direction and the acceleration is obvious, it is determined as a low impedance demand state; Introducing gait phase variable where 0 represents a full swing phase, 1 represents a full support phase, and a smooth transition Sigmoid function is designed between the two to avoid system oscillation caused by sudden changes in control parameters; Step 4.2 Construct a virtual admittance control model establishing an equation of desired dynamic relationship between the exoskeleton end effector and the wearer; the method will output the total disturbance estimation value of the extended state observer in step 3 as an input to the admittance model, is passed through the inertia matrix scaled to represent the interaction torque ; The admittance control law is defined as follows: ； in, The reference expected trajectory that the exoskeleton should follow, calculated by the admittance controller; This is the equilibrium position under no external force. For virtual inertia parameters; , These are the time-varying virtual damping coefficient and virtual stiffness coefficient, respectively; For the purpose of step 3 The calculated real-time human-computer interaction torque estimate; , They are respectively The first and second derivatives; Step 4.3 Design of variable parameter mapping rule based on disturbance observation and gait To achieve the adaptive effect of walking when people walk and stopping when people stop, the dynamic adjustment strategy of and is designed; the strategy depends on the size of gait phase and disturbance observation value at the same time: Designing stiffness adjustment strategies : ； wherein, is a minimum stiffness threshold; is a maximum stiffness threshold; in the support phase increases stiffness to provide load support; in the swing phase decreases stiffness to reduce motion resistance; Designing a damping adjustment strategy : Determining the strength of a person's movement intention using the amplitude of perturbation observations ​​ ； wherein, is the reference damping; is the adjustment factor; when the observer detects a large interaction torque, the algorithm automatically reduces the virtual damping , reducing the stickiness of the system, thus speeding up the response of the exoskeleton to human motion, eliminating the sense of asynchronization between man and machine; Step 4.4 Closed loop control command generation and execution the desired position and desired velocity as the tracking target for the underlying linear adaptive disturbance observer controller, i.e. the command reference in step 2 ​ The final execution logic of the control system is as follows: Step 4.4.1 Extended State Observer Real-Time Output Disturbance ; Step 4.4.2 The admittance controller computes the reference trajectory that is compliant with the human intent based on and the gait phase ; Step 4.4.3 will be With the position loop, the high disturbance rejection property of the LADRC itself is utilized again to drive the motor to track the trajectory precisely while compensating for the external friction and gravity.

Images