Dynamic walking exoskeleton control method and system based on human-machine complementary optimization

Abstract

The application provides a dynamic walking exoskeleton control method and system based on human-computer complementary optimization, comprising the following steps: defining the step height of the contact point, the normal contact force, constructing the nonlinear complementary problem constraint and model; defining the centroid position tracking of the contact point, the centroid linear velocity tracking, the foot yaw angle alignment, the frame attitude tracking, the base quaternion derivative regularization, the contact force regularization, the joint regularization, the swing height control, the contact control regularization, 9 types of motion task quantization targets, and establishing a multi-task objective function; coupling the static nonlinear complementary problem with the exoskeleton continuous dynamics to construct a coupled system; solving the optimization objective function by using a neural dynamics algorithm to obtain the optimal control input of the exoskeleton-human system; and mapping and converting the control input obtained by optimization into the driving instructions of the robot executor to complete the motion control closed loop. The application can more accurately adapt to the complex motion requirements under the human-robot strongly coupled constraint interaction.

Description

A Dynamic Walking Exoskeleton Control Method and System Based on Human-Machine Complementarity Optimization Technical Field

[0001] This invention relates to the technical field of optimized control of exoskeleton robots, specifically to a dynamic walking exoskeleton control method and system based on human-machine complementary optimization. More particularly, it relates to a dynamic walking exoskeleton control method and system based on model prediction and nonlinear complementary control. Background Technology

[0002] Exoskeletons are mechanical devices that assist human movement, providing support for people with mobility impairments and helping hemiplegic and paraplegic patients regain their ability to stand and walk. They can reduce muscle load on manual laborers, improve work efficiency in industrial and logistics settings, and enhance human motor skills, aiding in complex environments for work and training. However, current mainstream walking exoskeleton systems generally rely on predefined gait templates for control, severely limiting their user experience and safety.

[0003] The system crudely simplifies the complex foot-ground interaction process into a binary "either contact or separation" switch, completely ignoring the continuous force changes in the foot's sensation from heel to toe during walking. This directly leads to noticeable lag when walking and makes the user prone to slipping on wet or uneven surfaces due to the lack of force feedback. Therefore, a control method that can reproduce natural foot-ground interaction and provide precise control is a key breakthrough for achieving natural and stable walking with exoskeletons.

[0004] Chinese invention patent document CN202511037992X discloses an adaptive control method and related equipment for an exoskeleton robot. The method includes: acquiring motion and position information of each active joint in the exoskeleton robot; inputting the motion and position information into a pre-trained angle prediction model to obtain the target angle of each active joint; using the difference between the target angle and the actual angle output by the exoskeleton controller as the angle error; constructing a virtual control quantity based on the angle error; inputting the virtual control quantity and the motion parameters of the exoskeleton robot into an interference observer to generate an estimated interference; inputting the angle error, the virtual control quantity, and the estimated interference into an adaptive neural network controller to obtain the control torque of each active joint; and generating control signals for each active joint based on the control torque of each active joint.

[0005] Regarding the aforementioned technologies, the inventors believe that current control strategies for walking exoskeletons rely on predefined gait templates, and foot-ground contact behavior is often simplified into discrete state switching, which easily leads to control signal oscillations and poor adaptability. It is necessary to construct a continuous constraint model that integrates exoskeleton autonomy and foot-ground interaction, and to consider various constraints of strong coupling between humans and robots in real time, implementing a nonlinear complementary optimization control framework for the human-robot coupled system to address the above-mentioned shortcomings. Summary of the Invention

[0006] To address the shortcomings of existing technologies, the purpose of this invention is to provide a dynamic walking exoskeleton control method and system based on complementary constraints and optimization between humans and robots.

[0007] The dynamic walking exoskeleton control method based on human-machine complementary optimization provided by the present invention includes:

[0008] Problem modeling steps: Define the step height and normal contact force at the contact point between the exoskeleton and the ground; based on the distance-contact force complementarity condition and the velocity-contact force complementarity condition, construct a nonlinear complementary problem constraint and model that describes the physical rules of foot-ground contact.

[0009] Function establishment steps: Define the quantization objectives for nine types of motion tasks, including contact point centroid position tracking, centroid linear velocity tracking, foot yaw angle alignment, frame attitude tracking, base quaternion derivative regularization, contact force regularization, joint regularization, swing height control, and contact control regularization. Establish a multi-task objective function based on the quantization objectives of each task.

[0010] Constraint establishment steps: Couple the nonlinear complementary problem with the continuous dynamic equations of the exoskeleton to construct a coupled system model that simultaneously includes dynamic constraints describing system evolution, equality constraints describing kinematic relationships, and inequality constraints ensuring physical feasibility.

[0011] Problem-solving steps: Based on the multi-task objective function and the coupled system model, construct the model predictive control optimization problem, and use the neurodynamics algorithm to solve it in the rolling time domain to obtain the optimal control input sequence of the exoskeleton-human system in the future time domain;

[0012] Control input and execution steps: The instantaneous control inputs in the optimal control input sequence are mapped and converted into drive commands for the robot actuator to drive the exoskeleton movement, and the system status is collected in real time and fed back to the problem-solving steps to form closed-loop control.

[0013] Preferably, in the problem modeling step, the model of the nonlinear complementarity problem includes:

[0014] The distance-contact force complementary condition satisfies the following conditions: the contact point does not penetrate the ground and the normal contact force is non-negative, and the relaxed complementary constraint is satisfied. The expression is:

[0015]

[0016] in, Indicates the first Each contact point at time World coordinates Indicates the first Contact points Distance to the ground, Indicates the first Contact points The unit normal vector at the ground. Indicates the first Each contact point at time The ground contact force received, It is a pre-defined positive number;

[0017] The velocity-contact force complementary condition, including static friction constraints, is expressed as follows:

[0018]

[0019] in, Indicates the static friction coefficient. Indicates the first Contact points The set of tangential unit vectors at the ground;

[0020] And the friction-velocity complementary constraint, the expression is:

[0021]

[0022] in, This represents the non-negative relaxation parameter. Indicates the first Each contact point at time speed; This represents the constructor for a diagonal matrix, used to construct a diagonal matrix with elements of a vector as diagonal elements and the remaining positions set to 0.

[0023] The distance-contact force dynamic complementary condition is determined by constructing a variable parameter dynamic model. The forced complement terms converge, expressed as:

[0024]

[0025]

[0026] in, This represents a constant design parameter that controls the basic rate of convergence. This represents a time-varying function used to dynamically adjust the convergence rate, adapt to different contact states, and design parameters. Used to scale the convergence rate of the formula. Represents a monotonically increasing odd function, ensuring The direction of change is opposite to its own sign; It is a pre-defined positive number; It is the product of step height and normal contact force;

[0027] The dynamic complementary condition of velocity and contact force is achieved through continuous switching functions. Smooth the contact state and constrain the rate of change of the normal contact force, where, Indicates the scaling factor. Represents the hyperbolic secant function. It is represented as a continuous contact / off-ground state switching function.

[0028] Preferably, in the function establishment step, the multi-task objective function is a weighted sum of the cost functions of the nine types of motion tasks, and the cost functions include:

[0029] The task cost for the centroid location of the contact point is expressed as:

[0030]

[0031] in, This indicates the position of the center of mass at the point of contact between the two feet. Indicates the desired centroid location of the contact point. Denotes the weighted L2 norm. It is a weight vector; The cost function representing the centroid location of the contact point;

[0032] The task cost of the centroid linear velocity is expressed as:

[0033]

[0034] in, The linear momentum representing the center of mass of the exoskeleton. It is the expected centroid velocity. This represents the expected mass center line momentum. This represents the total mass of the exoskeleton robot and the human. Denotes the weighted L2 norm. It is a weight vector; The cost function representing the centroid linear velocity task;

[0035] The cost of foot yaw is expressed as:

[0036]

[0037] in, Indicates the current yaw angle of the left / right foot. Indicates the desired yaw angle for the left / right foot; Indicates the current yaw angle In the world coordinate system A unit vector formed on the xy plane such that it is perpendicular to the world coordinate system. The angle between the x-axis and the x-axis is equal to the desired yaw angle. ; Indicates the desired yaw angle In the world coordinate system A unit vector formed on the xy plane such that it is perpendicular to the world coordinate system. of The angle between the axes is equal to the desired yaw angle. ; Indicates and The transpose of a perpendicular unit vector. Indicates perpendicular to The vector, express transpose, Indicates the left foot. Indicates the right foot. Represents the L2 norm; The cost function representing the foot yaw task;

[0038] The cost of pose assignment in local coordinate systems is expressed as:

[0039]

[0040] in, This represents the deviation rotation matrix, which is the rotation required to transition from the desired attitude to the actual attitude. This means converting the deviation rotation matrix into a quaternion. Represents a unit quaternion; The cost function representing the pose task in the local coordinate system;

[0041] The cost of base quaternion derivative regularization is expressed as:

[0042]

[0043] in, This represents the quaternion rate of change of the base attitude with respect to time; This represents the expected quaternion rate of change of the base attitude with respect to time. The cost function for the base quaternion derivative regularization task is represented;

[0044] The cost of force regularization is expressed as:

[0045]

[0046] in, This represents the total number of contact points on a single foot of the exoskeleton. This represents the total contact force at all contact points; Indicates the first The expected force ratio at each contact point Indicates the first The expected force at each contact point, where j represents the index of all contact points on a single foot of the exoskeleton. Denotes the weighted L2 norm. It is a weight vector; The cost function representing the force regularization task;

[0047] The cost of joint regularization is expressed as:

[0048]

[0049] in, Indicates the actual angle of the joint. This indicates the desired angle that the joint will achieve. Indicates joint velocity. Indicates the adjustment parameter. Denotes the weighted L2 norm. It is a weight vector; The cost function represents the joint regularization task;

[0050] The task cost for foot swing height is expressed as:

[0051]

[0052] in, express The unit vector of the axis. express The unit vector of the axis. express The unit vector of the axis. Indicates contact point of The height of the direction, Indicates the desired foot swing height. Indicates the horizontal velocity at the point of contact. It is the exoskeleton in the control input. The velocity at each contact point; The cost function representing the foot-swinging height task;

[0053] The cost of contact control regularization is expressed as:

[0054]

[0055]

[0056] in, This indicates the contact point speed in the control input. Indicates the desired velocity at the contact point. This represents the rate of change of contact force in the control input. This represents the expected rate of change of contact force. This indicates the total number of contact points on a single foot of the exoskeleton. , Denotes the weighted L2 norm. , It is a weight vector; This represents the velocity tracking cost function for the contact control regularization task. This represents the force rate of change tracking cost function for the contact control regularization task.

[0057] Preferably, in the constraint establishment step, the coupled system model includes: dynamic constraints, equality constraints, and inequality constraints;

[0058] The dynamic constraints include:

[0059] Contact force change rate constraint:

[0060]

[0061] in, Indicates contact point The actual rate of change of contact force This indicates the rate of change of contact force in the control input; This means that the formula holds true for all contact points;

[0062] Contact point velocity constraint:

[0063]

[0064] in, Indicates contact point The actual speed Indicates the contact point in the control input. speed, This represents the speed transformation matrix, used to adjust the control input based on the contact point height. ;

[0065] System momentum rate of change constraint:

[0066]

[0067] in, This represents the rate of change of the total momentum of the human-exoskeleton system. This indicates the total number of contact points in the system. Represents the identity matrix. Indicates the first The location of each contact point Indicates the location of the system's centroid. This represents the cross product operation. Represents the cross product matrix. Represents the gravitational acceleration vector. Represents the gravity of the system. Represents the zero vector;

[0068] Center of gravity velocity constraint:

[0069]

[0070] in, Indicates the velocity of the system's center of gravity. Represents the linear momentum of the system;

[0071] Base speed constraint:

[0072]

[0073] in, Indicates the base in the world coordinate system The speed in the middle; Indicates the base in the local coordinate system The speed in the middle; Representing the local coordinate system To the world coordinate system The transformation matrix;

[0074] Base attitude change rate constraint:

[0075]

[0076] in, This represents the actual rate of change of attitude of the base. This represents the actual rate of change of attitude in the control input;

[0077] Joint velocity constraints:

[0078]

[0079] in, This represents the actual joint velocity. This indicates the joint speed in the control input;

[0080] The equality constraint is:

[0081]

[0082] in, Indicates contact point In the local coordinate system of the foot The position in the middle, Representing the local coordinate system of the foot To the world coordinate system The transformation matrix;

[0083] The inequality constraint is:

[0084]

[0085]

[0086]

[0087]

[0088]

[0089] in, Indicates a non-negative control boundary; This represents a non-negative control boundary.

[0090] Preferably, in the problem-solving step, the mathematical expression of the model predictive control optimization problem is:

[0091]

[0092]

[0093]

[0094]

[0095]

[0096]

[0097] in, Represents the state variables of the exoskeleton. Indicates control input, Represents the task weight vector. Represents the rate of change of a state variable. Represents the state transition function. Represents the constraint function. This represents the minimum height of the center of mass above the ground. The state function representing the height of the centroid along the z-axis. The upper and lower bound thresholds representing angular momentum. Represents the angular momentum of the system. express The unit vector of the axis. Indicates the position of the left foot's contact point. Indicates the position of the right foot contact point. This represents the upper and lower bound thresholds of the height difference between the contact points of the left and right feet along the z-axis.

[0098] The dynamic walking exoskeleton control system based on human-machine complementary optimization provided by the present invention includes:

[0099] Problem modeling module: Define the step height and normal contact force at the contact point between the exoskeleton and the ground, and construct a nonlinear complementary problem constraint and model describing the physical rules of foot-ground contact based on the distance-contact force complementarity condition and the velocity-contact force complementarity condition;

[0100] Function establishment module: Defines the quantization objectives for nine types of motion tasks, including contact point centroid position tracking, centroid linear velocity tracking, foot yaw angle alignment, frame attitude tracking, base quaternion derivative regularization, contact force regularization, joint regularization, swing height control, and contact control regularization. Based on the quantization objectives of each task, a multi-task objective function is established.

[0101] Constraint Establishment Module: Couples the nonlinear complementary problem with the continuous dynamic equations of the exoskeleton to construct a coupled system model that simultaneously includes dynamic constraints describing system evolution, equality constraints describing kinematic relationships, and inequality constraints ensuring physical feasibility.

[0102] Problem-solving module: Based on the multi-task objective function and the coupled system model, a model predictive control optimization problem is constructed, and a neurodynamics algorithm is used to solve it in the rolling time domain to obtain the optimal control input sequence of the exoskeleton-human system in the future time domain;

[0103] Control input and execution module: It maps the real-time control inputs in the optimal control input sequence into drive commands for the robot actuator, drives the exoskeleton to move, and collects system status feedback to the problem-solving module in real time to form closed-loop control.

[0104] Preferably, in the problem modeling module, the model of the nonlinear complementarity problem includes:

[0105] The distance-contact force complementary condition satisfies the following conditions: the contact point does not penetrate the ground and the normal contact force is non-negative, and the relaxed complementary constraint is satisfied. The expression is:

[0106]

[0107] in, Indicates the first Each contact point at time World coordinates Indicates the first Contact points Distance to the ground, Indicates the first Contact points The unit normal vector at the ground. Indicates the first Each contact point at time The ground contact force received, It is a pre-defined positive number;

[0108] The velocity-contact force complementary condition, including static friction constraints, is expressed as follows:

[0109]

[0110] in, Indicates the static friction coefficient. Indicates the first Contact points The set of tangential unit vectors at the ground;

[0111] And the friction-velocity complementary constraint, the expression is:

[0112]

[0113] in, This represents the non-negative relaxation parameter. Indicates the first Each contact point at time speed; This represents the constructor for a diagonal matrix, used to construct a diagonal matrix with elements of a vector as diagonal elements and the remaining positions set to 0.

[0114] The distance-contact force dynamic complementary condition is determined by constructing a variable parameter dynamic model. The forced complement terms converge, expressed as:

[0115]

[0116]

[0117] in, This represents a constant design parameter that controls the basic rate of convergence. This represents a time-varying function used to dynamically adjust the convergence rate, adapt to different contact states, and design parameters. Used to scale the convergence rate of the formula. Represents a monotonically increasing odd function, ensuring The direction of change is opposite to its own sign; It is a pre-defined positive number; It is the product of step height and normal contact force;

[0118] The dynamic complementary condition of velocity and contact force is achieved through continuous switching functions. Smooth the contact state and constrain the rate of change of the normal contact force, where, Indicates the scaling factor. Represents the hyperbolic secant function. It is represented as a continuous contact / off-ground state switching function.

[0119] Preferably, in the function establishment module, the multi-task objective function is a weighted sum of the cost functions of the nine types of motion tasks, and the cost functions include:

[0120] The task cost for the centroid location of the contact point is expressed as:

[0121]

[0122] in, This indicates the position of the center of mass at the point of contact between the two feet. Indicates the desired centroid location of the contact point. Denotes the weighted L2 norm. It is a weight vector; The cost function representing the centroid location of the contact point;

[0123] The task cost of the centroid linear velocity is expressed as:

[0124]

[0125] in, The linear momentum representing the center of mass of the exoskeleton. It is the expected centroid velocity. This represents the expected mass center line momentum. This represents the total mass of the exoskeleton robot and the human. Denotes the weighted L2 norm. It is a weight vector; The cost function representing the centroid linear velocity task;

[0126] The cost of foot yaw is expressed as:

[0127]

[0128] in, Indicates the current yaw angle of the left / right foot. Indicates the desired yaw angle for the left / right foot; Indicates the current yaw angle In the world coordinate system A unit vector formed on the xy plane such that it is perpendicular to the world coordinate system. The angle between the x-axis and the x-axis is equal to the desired yaw angle. ; Indicates the desired yaw angle In the world coordinate system A unit vector formed on the xy plane such that it is perpendicular to the world coordinate system. of The angle between the axes is equal to the desired yaw angle. ; Indicates and The transpose of a perpendicular unit vector. Indicates perpendicular to The vector, express transpose, Indicates the left foot. Indicates the right foot. Represents the L2 norm; The cost function representing the foot yaw task;

[0129] The cost of pose assignment in local coordinate systems is expressed as:

[0130]

[0131] in, This represents the deviation rotation matrix, which is the rotation required to transition from the desired attitude to the actual attitude. This means converting the deviation rotation matrix into a quaternion. Represents a unit quaternion; The cost function representing the pose task in the local coordinate system;

[0132] The cost of base quaternion derivative regularization is expressed as:

[0133]

[0134] in, This represents the quaternion rate of change of the base attitude with respect to time; This represents the expected quaternion rate of change of the base attitude with respect to time. The cost function for the base quaternion derivative regularization task is represented;

[0135] The cost of force regularization is expressed as:

[0136]

[0137] in, This represents the total number of contact points on a single foot of the exoskeleton. This represents the total contact force at all contact points; Indicates the first The expected force ratio at each contact point Indicates the first The expected force at each contact point, where j represents the index of all contact points on a single foot of the exoskeleton. Denotes the weighted L2 norm. It is a weight vector; The cost function representing the force regularization task;

[0138] The cost of joint regularization is expressed as:

[0139]

[0140] in, Indicates the actual angle of the joint. This indicates the desired angle that the joint will achieve. Indicates joint velocity. Indicates the adjustment parameter. Denotes the weighted L2 norm. It is a weight vector; The cost function represents the joint regularization task;

[0141] The task cost for foot swing height is expressed as:

[0142]

[0143] in, express The unit vector of the axis. express The unit vector of the axis. express The unit vector of the axis. Indicates contact point of The height of the direction, Indicates the desired foot swing height. Indicates the horizontal velocity at the point of contact. It is the exoskeleton in the control input. The velocity at each contact point; The cost function representing the foot-swinging height task;

[0144] The cost of contact control regularization is expressed as:

[0145]

[0146]

[0147] in, This indicates the contact point speed in the control input. Indicates the desired velocity at the contact point. This represents the rate of change of contact force in the control input. This represents the expected rate of change of contact force. This indicates the total number of contact points on a single foot of the exoskeleton. , Denotes the weighted L2 norm. , It is a weight vector; This represents the velocity tracking cost function for the contact control regularization task. This represents the force rate of change tracking cost function for the contact control regularization task.

[0148] Preferably, in the constraint establishment module, the coupled system model includes: dynamic constraints, equality constraints, and inequality constraints;

[0149] The dynamic constraints include:

[0150] Contact force change rate constraint:

[0151]

[0152] in, Indicates contact point The actual rate of change of contact force This indicates the rate of change of contact force in the control input; This means that the formula holds true for all contact points;

[0153] Contact point velocity constraint:

[0154]

[0155] in, Indicates contact point The actual speed Indicates the contact point in the control input. speed, This represents the speed transformation matrix, used to adjust the control input based on the contact point height. ;

[0156] System momentum rate of change constraint:

[0157]

[0158] in, This represents the rate of change of the total momentum of the human-exoskeleton system. This indicates the total number of contact points in the system. Represents the identity matrix. Indicates the first The location of each contact point Indicates the location of the system's centroid. This represents the cross product operation. Represents the cross product matrix. Represents the gravitational acceleration vector. Represents the gravity of the system. Represents the zero vector;

[0159] Center of gravity velocity constraint:

[0160]

[0161] in, Indicates the velocity of the system's center of gravity. Represents the linear momentum of the system;

[0162] Base speed constraint:

[0163]

[0164] in, Indicates the base in the world coordinate system The speed in the middle; Indicates the base in the local coordinate system The speed in the middle; Representing the local coordinate system To the world coordinate system The transformation matrix;

[0165] Base attitude change rate constraint:

[0166]

[0167] in, This represents the actual rate of change of attitude of the base. This represents the actual rate of change of attitude in the control input;

[0168] Joint velocity constraints:

[0169]

[0170] in, This represents the actual joint velocity. This indicates the joint speed in the control input;

[0171] The equality constraint is:

[0172]

[0173] in, Indicates contact point In the local coordinate system of the foot The position in the middle, Representing the local coordinate system of the foot To the world coordinate system The transformation matrix;

[0174] The inequality constraint is:

[0175]

[0176]

[0177]

[0178]

[0179]

[0180] in, Indicates a non-negative control boundary; This represents a non-negative control boundary.

[0181] Preferably, in the problem-solving module, the mathematical expression of the model predictive control optimization problem is:

[0182]

[0183]

[0184]

[0185]

[0186]

[0187]

[0188] in, Represents the state variables of the exoskeleton. Indicates control input, Represents the task weight vector. Represents the rate of change of a state variable. Represents the state transition function. Represents the constraint function. This represents the minimum height of the center of mass above the ground. The state function representing the height of the centroid along the z-axis. The upper and lower bound thresholds representing angular momentum. Represents the angular momentum of the system. express The unit vector of the axis. Indicates the position of the left foot's contact point. Indicates the position of the right foot contact point. This represents the upper and lower bound thresholds of the height difference between the contact points of the left and right feet along the z-axis.

[0189] Compared with the prior art, the present invention has the following beneficial effects:

[0190] 1. This invention embeds the physical rules of foot-to-ground contact / non-contact with the moving target into the same nonlinear complementary optimization problem, overcoming the limitation of traditional MPC in directly handling "multi-contact mixed states";

[0191] 2. This invention addresses the difficulty of achieving stable tracking and balance in dynamic environments for exoskeleton robots by proposing a unified cost framework for multi-task weighted optimization. This framework integrates multi-dimensional tasks such as contact point position, momentum, and coordinate system attitude into a single optimization problem in the form of a weighted cost function, replacing the "sequential control of multiple tasks" logic of existing technologies. This framework can more accurately adapt to the complex motion requirements of humanoid robots.

[0192] 3. Existing technologies typically impose loose constraints on momentum / angular momentum. This invention uses generalized momentum (linear momentum + angular momentum) as both a cost function term and a hard constraint term, achieving both precise momentum tracking and safety assurance through dual control. This more strictly suppresses the risk of imbalance in exoskeleton robots and improves motion stability.

[0193] 4. Existing technologies often overlook the problem of abrupt changes in control inputs, which can easily lead to motor vibration and mechanical wear. This invention adds regularization costs to all control inputs such as contact force, position, base posture, and joints, which can effectively avoid abrupt changes in inputs and improve the smoothness of motion and mechanical life.

[0194] 5. Existing technologies often customize control logic for single tasks, while the "state-control-constraint" unified MPC optimization framework of this invention can quickly adapt to different tasks (such as exoskeletons going up and down stairs, walking on complex terrain) by adjusting task weights / constraints, thus reducing cross-scene development costs. Attached Figure Description

[0195] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0196] Figure 1 is a block diagram illustrating a dynamic walking exoskeleton control method based on human-machine complementary optimization according to the present invention;

[0197] Figure 2 shows the autonomous control framework based on MPC, dynamic nonlinear complementary conditions, and neurodynamic nonlinear solver involved in this invention. Detailed Implementation

[0198] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0199] Example

[0200] As shown in Figure 1, this invention provides a dynamic walking exoskeleton control method based on human-machine complementary optimization, comprising the following steps:

[0201] Problem modeling steps: Define the step height and normal contact force at the contact point, and construct the constraints and model for the nonlinear complementary problem;

[0202] Function establishment steps: Define the quantitative objectives for nine types of motion tasks, including contact point centroid position tracking, centroid linear velocity tracking, foot yaw angle alignment, frame attitude tracking, base quaternion derivative regularization, contact force regularization, joint regularization, swing height control, and contact control regularization, and establish a multi-task objective function;

[0203] Constraint establishment steps: Couple the static nonlinear complementary problem with the continuous dynamics of the exoskeleton to construct a coupled system that includes "dynamic constraints, equality constraints, and inequality constraints";

[0204] Problem-solving steps: Use a neurodynamics algorithm to solve the objective function and obtain the optimal control input for the exoskeleton-human system;

[0205] Control input and execution steps: The optimized control input is mapped and converted into drive instructions for the robot actuator to complete the motion control closed loop.

[0206] Preferably, in the problem modeling step, a dynamic nonlinear complementary system is used to model the multi-contact interactions.

[0207] Definition 1: Given a vector The nonlinear complementarity problem is to find a vector that satisfies the following system of equations and inequalities. :

[0208]

[0209] in, , ; Represents an arbitrary vector; Let m represent the set of m-dimensional real vectors.

[0210] 1) Complementary conditions:

[0211] ① Distance and contact force complement each other.

[0212] On a rigid, flat surface, the point of contact should not penetrate the walking surface, that is:

[0213]

[0214] in, Indicates the first Each contact point at time World coordinates Indicates the distance from the point of contact to the ground; Represents a set of 3-dimensional real vectors; Represents the set of real numbers;

[0215] Contact force The normal component is non-negative:

[0216]

[0217] in, Indicates the first Contact points The unit normal vector at the ground. Indicates the first The ground contact force experienced at each contact point.

[0218] reaction force The value of is only non-zero when the contact point is in contact with the traveling surface. Therefore, it can be expressed as the following distance complementary constraint:

[0219]

[0220] in, Indicates the distance from the point of contact to the ground. Indicates the first Contact points The unit normal vector at the ground. Indicates the first The ground contact force at each contact point Represents a very small positive number; Represent the set of positive real numbers;

[0221] Using inequality conditions as relaxation complements and a very small positive number as the upper bound of the product can avoid constraining Jacobian singularity and causing numerical problems.

[0222] ② Speed ​​and contact force complement each other.

[0223] When the contact force is insufficient to overcome static friction:

[0224]

[0225] in, Indicates the static friction coefficient. Indicates the first Contact points The set of tangential unit vectors at the ground.

[0226] By relaxation parameters The product range of the planar tangential velocity and tangential contact force at the contact point is limited, creating friction-velocity complementarity. When the contact point is in contact with the ground, ideally the tangential velocity should be 0, and the tangential force should be provided by static friction; when the contact point leaves the ground, the tangential force should also be 0, satisfying the constraint, and the velocity is not limited by force; during the transitional state of contact switching, through... Allowing for small velocity-force products avoids numerical oscillations during contact switching, ensuring the solvability of the optimization problem. The formula is:

[0227]

[0228] in, This represents the non-negative relaxation parameter. Indicates the first Each contact point at time speed; This represents the constructor for a diagonal matrix, used to construct a diagonal matrix with elements of a vector as diagonal elements and the remaining positions set to 0.

[0229] 2) Dynamic forced complementarity:

[0230] ①Distance-contact force dynamic complementarity.

[0231] Complementarity constraints can be implemented using neurodynamics, forcing complementary separation-contact forces to achieve convergence. Let:

[0232]

[0233] in, It is the contact point step height (vertical distance from the ground); It is the normal contact force at the point of contact; therefore It is the product of "step height × normal contact force", and complementary constraints require... (The step height is 0 when in contact and the force is 0 when not in contact).

[0234] Based on neurodynamic design experience, to make To approach 0, its time derivative needs to be negative (i.e., ... ),let (decreases over time). Therefore, a variable-parameter dynamic model was constructed:

[0235]

[0236] in, This represents a constant design parameter that controls the base rate of convergence. This represents a time-varying function used to dynamically adjust the convergence rate, adapt to different contact states, and design parameters. Used to scale the convergence rate of the formula. Odd functions that are monotonically increasing, such as hyperbolic tangent. ,make sure The direction of change is opposite to its own sign; It represents a small positive number.

[0237] In this model, to improve its practicality, two adjustments were made to the dynamic model: a. Constraint relaxation: Equations were changed to inequalities, resulting in faster convergence and better meeting real-time control requirements; b. Addition of a relaxation term: A small positive number was introduced. The purpose of further relaxing the constraints is to expand the feasible region, avoid the lack of solutions due to overly strict constraints during numerical solution, and improve the robustness of the algorithm.

[0238] ② Dynamic complementarity of velocity and contact force.

[0239] The following constraint is applied to the derivative of the contact force. When the exoskeleton contacts the ground, You can take freely Any value within the range. When the exoskeleton is not in contact with the ground, the control input causes the contact force to decrease exponentially, the rate of which depends on the positive definite control gain. .

[0240]

[0241]

[0242] in, Indicates the non-negative control boundary. The derivative of the contact force, Indicates the distance from the point of contact to the ground. Indicates positive definite control gain. Indicates the first The ground contact force at each contact point; Represents: a set of 3×3 real number matrices.

[0243] To avoid sudden changes in contact force caused by direct switching of discrete bivariate functions, a hyperbolic secant function is introduced to transform the discrete state of "contact / lift-off" into a continuous value between 0 and 1, let:

[0244]

[0245] in, Indicates the scaling factor. Indicates the distance from the point of contact to the ground. Represents the hyperbolic secant function. It is represented as a continuous "contact / off-ground" state switching function.

[0246] By simplifying the lower bound of the above equation, it can be written as an inequality:

[0247]

[0248] Among them, when At that time, ;when When the boundary overlaps, it equals .

[0249] Considering contact force The tangential force is constrained by static friction. Therefore, as long as the normal contact force decreases to zero, the tangential force will naturally also decrease to zero to satisfy the friction constraint. Therefore, it is only necessary to constrain the rate of change of the normal contact force, further simplifying the constraint:

[0250]

[0251] in, express The unit vector of the axis. express Positive definite control gain in the axial direction.

[0252] Preferably, in the function establishment step, a nonlinear trajectory optimization problem combining contact complementarity and moving target is formulated within a unified Model Predictive Control (MPC) framework, and the complementary constraints are encoded into the MPC formula to solve for a dynamically feasible walking trajectory.

[0253] 1) Definition of motor task:

[0254] To enable an exoskeleton robot to move to a desired location, a task is designed in Cartesian space.

[0255] ① Contact point centroid location task:

[0256] The goal is to minimize the error between the centroid of the foot contact point of the exoskeleton robot and its desired position in the absolute coordinate system.

[0257]

[0258] in, Indicates the position of the center of mass of the point of contact between the two feet ( Indicates the position of the left foot's contact point. (This indicates the position of the right foot's contact point, considering only the x and y directions of the horizontal plane). Indicates the desired centroid location of the contact point. Denotes the weighted L2 norm. It is a weight vector; The cost function represents the task of determining the centroid location at the contact point.

[0259] ② Center of mass linear velocity task:

[0260] The goal is to minimize the error between the velocity of the exoskeleton robot's center of gravity and the desired center of gravity velocity.

[0261]

[0262] in, The linear momentum representing the center of mass of the exoskeleton. It is the expected centroid velocity. This represents the expected mass center line momentum. This represents the total mass of the exoskeleton robot and the human. Denotes the weighted L2 norm. It is a weight vector; The cost function represents the task of achieving the linear velocity of the center of mass.

[0263] ③ Foot yaw task:

[0264] The goal is to minimize the error between the yaw angle of the exoskeleton robot's left / right foot and the desired yaw angle.

[0265]

[0266] in, Indicates the current yaw angle of the left / right foot. Indicates the desired yaw angle for the left / right foot; Indicates the current yaw angle In the world coordinate system A unit vector formed on the xy plane such that it is perpendicular to the world coordinate system. The angle between the x-axis and the x-axis is equal to the desired yaw angle. ; Indicates the desired yaw angle In the world coordinate system A unit vector formed on the xy plane such that it is perpendicular to the world coordinate system. of The angle between the axes is equal to the desired yaw angle. ; Indicates and The transpose of a perpendicular unit vector. Indicates perpendicular to The vector, express transpose, Indicates the left foot. Indicates the right foot. Represents the L2 norm; The cost function represents the foot yaw task.

[0267] This task does not prevent the feet from rolling and pitching during the swing phase.

[0268] ④ Local coordinate system attitude task:

[0269] The goal is to minimize the error between the pose of the exoskeleton robot in the specified local coordinate system and the desired pose.

[0270]

[0271] in, Indicates the local coordinate system in the world coordinate system The desired rotation matrix (i.e., desired pose) in the matrix. Indicates the local coordinate system in the world coordinate system The actual rotation matrix (i.e., the actual attitude) in the matrix. This represents the deviation rotation matrix, which is the rotation required to transition from the desired attitude to the actual attitude. This means converting the deviation rotation matrix into a quaternion. It represents a unit quaternion (i.e., a standard orientation without rotation). Represents the L2 norm; The cost function represents the attitude task in the local coordinate system.

[0272] ⑤ Base quaternion derivative regularization task:

[0273] The goal is to minimize the error between the quaternion rate of change corresponding to the pose of the exoskeleton robot's base (exoskeleton torso) and the desired quaternion rate of change.

[0274]

[0275] in, This represents the quaternion rate of change of the base attitude with respect to time; This represents the expected quaternion rate of change of the base attitude with respect to time. The cost function for the base quaternion derivative regularization task is represented; Represents a set of 4-dimensional real vectors.

[0276] ⑥ Force regularization task:

[0277] The goal is to minimize the error between the actual force ratio and the desired force ratio at each contact point of the exoskeleton robot.

[0278]

[0279] in, Indicates the first The actual contact force at each contact point; This indicates the total number of contact points on a single foot (left / right foot) of the exoskeleton; Indicates the left foot. Indicates the right foot; This represents the total contact force at all contact points; Indicates the first The expected force ratio at each contact point Indicates the first The expected force at each contact point, where j represents the index of all contact points on a single foot (left / right) of the exoskeleton. Denotes the weighted L2 norm. It is a weight vector; The cost function representing the force regularization task;

[0280] ⑦ Joint regularization task:

[0281] The goal is to minimize the error between the joint angles of the exoskeleton robot and the desired angles.

[0282]

[0283] in, Indicates the actual angle of the joint. This indicates the desired angle that the joint will achieve. Indicates joint velocity. This represents the adjustment parameters (positive semi-definite matrix, which controls the speed at which the joint converges to the desired posture). Indicates the number of joints. Denotes the weighted L2 norm. It is a weight vector; The cost function represents the joint regularization task; Represents the set of n-dimensional real vectors; Represents the set of n×n dimensional real matrices;

[0284] ⑧ Foot swing height task:

[0285] The goal is to minimize the error between the height of the exoskeleton robot's swinging foot and the desired height.

[0286]

[0287] in, express The unit vector of the axis. express The unit vector of the axis. express The unit vector of the axis. Indicates contact point of The height of the direction, Indicates the desired foot swing height. Indicates the horizontal velocity at the point of contact. It is the exoskeleton in the control input. The velocity at each contact point; The cost function representing the foot-swinging height task;

[0288] ⑨ Contact control regularization task:

[0289] The goal is to minimize the error between the rate of change of velocity and force at the contact point of the exoskeleton robot and the desired rate of change of velocity and force.

[0290]

[0291]

[0292] in, This indicates the contact point speed in the control input. Indicates the desired velocity at the contact point. This represents the rate of change of contact force in the control input. This represents the expected rate of change of contact force. This indicates the total number of contact points on a single foot (left / right) of the exoskeleton. , Denotes the weighted L2 norm. , It is a weight vector; This represents the velocity tracking cost function for the contact control regularization task. This represents the force rate of change tracking cost function for the contact control regularization task.

[0293] Preferably, in the constraint establishment step, the "human-exoskeleton interaction" is modeled as a differential-algebraic system that can jointly regulate body dynamics and ground interaction. This system consists of the following components:

[0294] 1) Dynamic constraints:

[0295] It describes the changes in contact force and contact point location, as well as the center of mass momentum and exoskeleton state. Contact force and contact points By controlling the input and Updates are being performed, and system momentum... Center of mass Base position ,direction and exoskeleton joint angles It evolves according to whole-body dynamics.

[0296] ①Contact force change rate constraint:

[0297]

[0298] in, Indicates contact point The actual rate of change of contact force This indicates the rate of change of contact force in the control input; This means that the formula holds true for all contact points.

[0299] ②Contact point velocity constraint:

[0300]

[0301] in, Indicates contact point The actual speed Indicates the contact point in the control input. speed, This represents the speed transformation matrix, used to adjust the control input based on the contact point height. .

[0302] ③ Constraints on the rate of change of system momentum:

[0303]

[0304] in, This represents the rate of change of the total momentum of the "human-exoskeleton" system. This indicates the total number of contact points in the system. Represents the identity matrix. Indicates the first The location of each contact point Indicates the location of the system's centroid. This represents the cross product operation. Represents the cross product matrix. Indicates the first The ground contact force at each contact point Indicates the total mass of the system. Represents the gravitational acceleration vector. Represents the gravity of the system. This represents the zero vector.

[0305] ④ Center of gravity velocity constraint:

[0306]

[0307] in, Indicates the velocity of the system's center of gravity. It represents the linear momentum of the system. This indicates the total mass of the system.

[0308] ⑤ Base speed constraint:

[0309]

[0310] in, Indicates the base in the world coordinate system The speed in the middle; Indicates the base in the local coordinate system The speed in the middle; Representing the local coordinate system To the world coordinate system The transformation matrix.

[0311] ⑥ Base attitude change rate constraint:

[0312]

[0313] in, This represents the rate of change of the actual attitude (quaternion) of the base. This represents the actual rate of change of attitude (quaternion) in the control input.

[0314] ⑦ Joint velocity constraints:

[0315]

[0316] in, This represents the actual joint velocity. This indicates the joint speed in the control input. Indicates the number of joints.

[0317] 2) Equality constraints:

[0318]

[0319] in, Indicates contact point In the world coordinate system The position in the middle, Indicates contact point In the local coordinate system of the foot The position in the middle, Representing the local coordinate system of the foot To the world coordinate system The transformation matrix.

[0320] 3) Inequality constraints:

[0321] These constraints ensure physical feasibility, including:

[0322]

[0323] in, Indicates the first Contact points The unit normal vector at the ground. Indicates the first The ground contact force experienced at each contact point.

[0324]

[0325] in, Indicates the first Contact points The set of tangential unit vectors at the ground. Indicates the first The ground contact force at each contact point Indicates the static friction coefficient. Indicates the first Contact points The unit normal vector of the ground at that location.

[0326]

[0327] in, Indicates the non-negative control boundary. It is the exoskeleton in the control input. The speed of each contact point.

[0328]

[0329] in, Indicates the non-negative control boundary. This represents the rate of change of contact force in the control input.

[0330]

[0331] in, This indicates the distance from the point of contact to the ground.

[0332] Preferably, in the problem-solving steps, based on the above tasks and constraints, the following mathematical optimization problem is formulated:

[0333]

[0334]

[0335] See the above dynamic constraints.

[0336] See the above equality constraints and inequality constraints.

[0337]

[0338]

[0339]

[0340] in, Represents the state variables of the exoskeleton. Indicates control input, Represents the task weight vector. Represents the rate of change of a state variable. Represents the state transition function. Represents the constraint function. This represents the minimum height of the center of mass above the ground. The state function representing the height of the centroid along the z-axis. The upper and lower bound thresholds representing angular momentum. Represents the angular momentum of the system. express The unit vector of the axis. Indicates the position of the left foot's contact point. Indicates the position of the right foot contact point. This represents the upper and lower bound thresholds of the height difference between the contact points of the left and right feet along the z-axis.

[0341] Based on the objective function, a neurodynamic algorithm is used to solve the objective function, obtaining the optimal control input for the exoskeleton-human system. The planner is repeatedly invoked using the rolling time-domain principle, with the previous solution used to warm-start the current optimization each time, thus achieving online replanning.

[0342] Preferably, in the control input and execution steps, the optimized control input is mapped and converted into drive commands for the exoskeleton robot actuator. Sensors collect the robot's actual motion state in real time and provide feedback, forming a motion control closed loop. The planner is repeatedly invoked using the rolling time-domain principle, with the previous solution used to hot-start the current optimization each time, thus achieving online replanning.

[0343] This invention also provides a dynamic walking exoskeleton control system based on human-machine complementary optimization, comprising the following modules:

[0344] Problem modeling module: Defines the step height and normal contact force at the contact point, and constructs the constraints and model for the nonlinear complementary problem;

[0345] Function creation module: Define the quantitative objectives for nine types of motion tasks, including contact point centroid position tracking, centroid linear velocity tracking, foot yaw angle alignment, frame attitude tracking, base quaternion derivative regularization, contact force regularization, joint regularization, swing height control, and contact control regularization, and establish a multi-task objective function;

[0346] Constraint Establishment Module: Couples the static nonlinear complementary problem with the continuous dynamics of the exoskeleton to construct a coupled system that includes "dynamic constraints, equality constraints, and inequality constraints";

[0347] Problem-solving module: Employs a neurodynamics algorithm to solve the objective function and obtain the optimal control input for the exoskeleton-human system. By repeatedly calling the planner using the rolling time-domain principle, the current optimization is restarted using the previous solution each time, thus achieving online replanning.

[0348] Control Input and Execution Module: Converts the optimized control input mapping into drive commands for the robot actuator, completing the motion control closed loop.

[0349] As shown in Figure 2, the specific process of this invention is as follows:

[0350] 1. Define the walking task and physical constraints of the exoskeleton; provide contact complementary constraints based on the physical rules of foot on the ground / off the ground to limit the boundaries for subsequent optimization;

[0351] 2. Based on the task and constraints, MPC plans the optimal control input sequence for the next T time steps using a rolling time domain approach. This information is passed to the upper-level neurodynamic nonlinear optimization solver.

[0352] 3. Optimal control input obtained through optimization After entering the neurodynamics solver and completing fine-tuning, the force is passed to the PD control + Jacobian matrix module. This module maps the contact force into joint torques of the exoskeleton using the Jacobian matrix. Meanwhile, PD control is introduced to ensure the stability of joint trajectory tracking;

[0353] 4. Real-time closed-loop joint torque After driving the exoskeleton to move, the system status (center of gravity position) is collected in real time using the center of mass dynamics. Joint angle Walking speed These states are fed back to two modules: the MPC module, which serves as the "current state" for the next round of rolling optimization; and the "human-exoskeleton collaboration" module, which corrects subsequent control objectives.

[0354] 5. Real-time acquisition of motion data from the "exoskeleton-human" system is fed back to the PD control module to form a complete closed-loop control flow.

[0355] Those skilled in the art will understand that, in addition to implementing the system, apparatus, and their modules provided by this invention in purely computer-readable program code, the same program can be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system, apparatus, and their modules provided by this invention can be considered a hardware component, and the modules included therein for implementing various programs can also be considered structures within the hardware component; alternatively, modules for implementing various functions can be considered both software programs implementing the method and structures within the hardware component.

[0356] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A dynamic walking exoskeleton control method based on human-machine complementary optimization, characterized in that, include: Problem modeling steps: Define the step height and normal contact force at the contact point between the exoskeleton and the ground. Based on the distance-contact force complementarity condition and the velocity-contact force complementarity condition, construct a nonlinear complementary problem constraint and model describing the physical rules of foot-ground contact. Function establishment steps: Define quantified objectives for nine types of motion tasks, including contact point centroid position tracking, centroid linear velocity tracking, foot yaw angle alignment, frame attitude tracking, base quaternion derivative regularization, contact force regularization, joint regularization, swing height control, and contact control regularization. Establish a multi-task objective function based on the quantified objectives of each task. Constraint establishment steps: Integrate the aforementioned nonlinear complementary problem with the continuous dynamic equations of the exoskeleton. The system is coupled to construct a coupled system model that simultaneously includes dynamic constraints describing system evolution, equality constraints describing kinematic relationships, and inequality constraints ensuring physical feasibility. The problem-solving steps are as follows: Based on the multi-task objective function and the coupled system model, a model predictive control optimization problem is constructed and solved in the rolling time domain using a neurodynamics algorithm to obtain the optimal control input sequence for the exoskeleton-human system in the future time domain. The control input and execution steps involve mapping the instantaneous control inputs in the optimal control input sequence into drive commands for the robot actuators, driving the exoskeleton's movement, and real-time system state feedback to the problem-solving steps, forming a closed-loop control. In the problem modeling steps, the model of the nonlinear complementary problem includes: the distance-contact force complementarity condition, which satisfies that the contact point does not penetrate the ground and the normal contact force is non-negative, and also satisfies the relaxed complementarity constraint, expressed as: in, Indicates the first Each contact point at time World coordinates Indicates the first Contact points Distance to the ground, Indicates the first Contact points The unit normal vector at the ground. Indicates the first Each contact point at time The ground contact force received, It is a preset positive number; the velocity-contact force complementary condition, including static friction constraints, is expressed as: in, Indicates the static friction coefficient. Indicates the first Contact points The set of tangential unit vectors on the ground; and the friction-velocity complementary constraint, expressed as: in, This represents the non-negative relaxation parameter. Indicates the first Each contact point at time speed; This represents the constructor for a diagonal matrix, used to construct a diagonal matrix with vector elements as diagonal elements and the remaining positions set to 0; the distance-contact force dynamic complementary condition is constructed by creating a variable parameter dynamic model. The forced complement terms converge, expressed as: in, This represents a constant design parameter that controls the basic rate of convergence. This represents a time-varying function used to dynamically adjust the convergence rate, adapt to different contact states, and design parameters. Used to scale the convergence rate of the formula. Represents a monotonically increasing odd function, ensuring The direction of change is opposite to its own sign; It is a pre-defined positive number; It is the product of step height and normal contact force; the velocity-contact force dynamic complementarity condition is obtained through continuously switching functions. Smooth the contact state and constrain the rate of change of the normal contact force, where, Indicates the scaling factor. Represents the hyperbolic secant function. It is represented as a continuous contact / off-ground state switching function.

2. The dynamic walking exoskeleton control method based on human-machine complementary optimization according to claim 1, characterized in that, In the function establishment step, the multi-task objective function is a weighted sum of the cost functions of the nine types of motion tasks. The cost function includes the task cost for the contact point centroid position, expressed as: in, This indicates the position of the center of mass at the point of contact between the two feet. Indicates the desired centroid location of the contact point. Denotes the weighted L2 norm. It is a weight vector; The cost function for the task of determining the centroid position at the contact point; the cost function for the task of determining the centroid linear velocity, expressed as: in, The linear momentum representing the center of mass of the exoskeleton. It is the expected centroid velocity. This represents the expected mass center line momentum. This represents the total mass of the exoskeleton robot and the human. Denotes the weighted L2 norm. It is a weight vector; The cost function for the centroid linear velocity task; the cost function for the foot yaw task, expressed as: in, Indicates the current yaw angle of the left / right foot. Indicates the desired yaw angle for the left / right foot; Indicates the current yaw angle In the world coordinate system A unit vector formed on the xy plane such that it is perpendicular to the world coordinate system. The angle between the x-axis and the x-axis is equal to the desired yaw angle. ； Indicates the desired yaw angle In the world coordinate system A unit vector formed on the xy plane such that it is perpendicular to the world coordinate system. of The angle between the axes is equal to the desired yaw angle. ； Indicates and The transpose of a perpendicular unit vector. Indicates perpendicular to The vector, express transpose, Indicates the left foot. Indicates the right foot. Represents the L2 norm; The cost function represents the foot yaw task; the local coordinate system attitude task cost is expressed as: in, This represents the deviation rotation matrix, which is the rotation required to transition from the desired attitude to the actual attitude. This means converting the deviation rotation matrix into a quaternion. Represents a unit quaternion; The cost function for the local coordinate system pose task is represented by the base quaternion derivative regularization cost, expressed as: in, This represents the quaternion rate of change of the base attitude with respect to time; This represents the expected quaternion rate of change of the base attitude with respect to time. The cost function for the base quaternion derivative regularization task is expressed as: in, This represents the total number of contact points on a single foot of the exoskeleton. This represents the total contact force at all contact points; Indicates the first The expected force ratio at each contact point Indicates the first The expected force at each contact point, where j represents the index of all contact points on a single foot of the exoskeleton. Denotes the weighted L2 norm. It is a weight vector; The cost function for force regularization is given by: The cost function for joint regularization is given by: in, Indicates the actual angle of the joint. This indicates the desired angle that the joint will achieve. Indicates joint velocity. Indicates the adjustment parameter. Denotes the weighted L2 norm. It is a weight vector; The cost function for the joint regularization task is represented by the foot swing height task cost, expressed as: in, express The unit vector of the axis. express The unit vector of the axis. express The unit vector of the axis. Indicates contact point of The height of the direction, Indicates the desired foot swing height. Indicates the horizontal velocity at the point of contact. It is the exoskeleton in the control input. The velocity at each contact point; The cost function for the foot swing height task is represented by the contact control regularization task cost, expressed as: in, This indicates the contact point speed in the control input. Indicates the desired velocity at the contact point. This represents the rate of change of contact force in the control input. This represents the expected rate of change of contact force. This indicates the total number of contact points on a single foot of the exoskeleton. 、 Denotes the weighted L2 norm. 、 It is a weight vector; This represents the velocity tracking cost function for the contact control regularization task. This represents the force rate of change tracking cost function for the contact control regularization task.

3. The dynamic walking exoskeleton control method based on human-machine complementary optimization according to claim 2, characterized in that, In the constraint establishment step, the coupled system model includes: dynamic constraints, equality constraints, and inequality constraints; the dynamic constraints include: contact force change rate constraints. in, Indicates contact point The actual rate of change of contact force This indicates the rate of change of contact force in the control input; This means that the formula holds true for all contact points; contact point velocity constraint: in, Indicates contact point The actual speed Indicates the contact point in the control input. speed, This represents the speed transformation matrix, used to adjust the control input based on the contact point height. System momentum change rate constraint: in, This represents the rate of change of the total momentum of the human-exoskeleton system. This indicates the total number of contact points in the system. Represents the identity matrix. Indicates the first The location of each contact point Indicates the location of the system's centroid. This represents the cross product operation. Represents the cross product matrix. Represents the gravitational acceleration vector. Represents the gravity of the system. Represents the zero vector; centroid velocity constraint: in, Indicates the velocity of the system's center of gravity. Represents the linear momentum of the system; base velocity constraint: in, Indicates the base in the world coordinate system The speed in the middle; Indicates the base in the local coordinate system The speed in the middle; Representing the local coordinate system To the world coordinate system Transformation matrix; Base attitude change rate constraint: in, This represents the actual rate of change of attitude of the base. Indicates the actual rate of change of attitude in the control input; joint velocity constraints: in, This represents the actual joint velocity. This represents the joint speed in the control input; the equality constraint is: in, Indicates contact point In the local coordinate system of the foot The position in the middle, Representing the local coordinate system of the foot To the world coordinate system The transformation matrix; the inequality constraints are: in, Indicates a non-negative control boundary; This represents a non-negative control boundary.

4. The dynamic walking exoskeleton control method based on human-machine complementary optimization according to claim 3, characterized in that, In the problem-solving steps, the mathematical expression of the model predictive control optimization problem is as follows: in, Represents the state variables of the exoskeleton. Indicates control input, Represents the task weight vector. Represents the rate of change of a state variable. Represents the state transition function. Represents the constraint function. This represents the minimum height of the center of mass above the ground. The state function representing the height of the centroid along the z-axis. The upper and lower bound thresholds representing angular momentum. Represents the angular momentum of the system. express The unit vector of the axis. Indicates the position of the left foot's contact point. Indicates the position of the right foot contact point. This represents the upper and lower bound thresholds of the height difference between the contact points of the left and right feet along the z-axis.

5. A dynamic walking exoskeleton control system based on human-machine complementary optimization, characterized in that, include: Problem Modeling Module: Defines the step height and normal contact force at the contact point between the exoskeleton and the ground. Based on the distance-contact force complementarity condition and the velocity-contact force complementarity condition, it constructs a nonlinear complementary problem constraint and model describing the physical rules of foot-ground contact. Function Establishment Module: Defines quantified objectives for nine types of motion tasks, including contact point centroid position tracking, centroid linear velocity tracking, foot yaw angle alignment, frame attitude tracking, base quaternion derivative regularization, contact force regularization, joint regularization, swing height control, and contact control regularization. Based on the quantified objectives of each task, it establishes a multi-task objective function. Constraint Establishment Module: Integrates the aforementioned nonlinear complementary problem with the continuous dynamic equations of the exoskeleton. The system is coupled to construct a coupled system model that simultaneously includes dynamic constraints describing system evolution, equality constraints describing kinematic relationships, and inequality constraints ensuring physical feasibility. The problem-solving module, based on the multi-task objective function and the coupled system model, constructs a model predictive control optimization problem and uses a neurodynamics algorithm for rolling time-domain solution to obtain the optimal control input sequence for the exoskeleton-human system in the future time domain. The control input and execution module maps the instantaneous control inputs in the optimal control input sequence into drive commands for the robot actuators, driving the exoskeleton's movement, and collects system state feedback in real time to the problem-solving module, forming a closed-loop control. In the problem modeling module, the model of the nonlinear complementary problem includes: the distance-contact force complementarity condition, which satisfies that the contact point does not penetrate the ground and the normal contact force is non-negative, and also satisfies the relaxed complementarity constraint, expressed as: in, Indicates the first Each contact point at time World coordinates Indicates the first Contact points Distance to the ground, Indicates the first Contact points The unit normal vector at the ground. Indicates the first Each contact point at time The ground contact force received, It is a preset positive number; the velocity-contact force complementary condition, including static friction constraints, is expressed as: in, Indicates the static friction coefficient. Indicates the first Contact points The set of tangential unit vectors on the ground; and the friction-velocity complementary constraint, expressed as: in, This represents the non-negative relaxation parameter. Indicates the first Each contact point at time speed; This represents the constructor for a diagonal matrix, used to construct a diagonal matrix with vector elements as diagonal elements and the remaining positions set to 0; the distance-contact force dynamic complementary condition is constructed by creating a variable parameter dynamic model. The forced complement terms converge, expressed as: in, This represents a constant design parameter that controls the basic rate of convergence. This represents a time-varying function used to dynamically adjust the convergence rate, adapt to different contact states, and design parameters. Used to scale the convergence rate of the formula. Represents a monotonically increasing odd function, ensuring The direction of change is opposite to its own sign; It is a pre-defined positive number; It is the product of step height and normal contact force; the velocity-contact force dynamic complementarity condition is obtained through continuously switching functions. Smooth the contact state and constrain the rate of change of the normal contact force, where, Indicates the scaling factor. Represents the hyperbolic secant function. It is represented as a continuous contact / off-ground state switching function.

6. The dynamic walking exoskeleton control system based on human-machine complementary optimization according to claim 5, characterized in that, In the function establishment module, the multi-task objective function is a weighted sum of the cost functions of the nine types of motion tasks. The cost function includes the task cost for the contact point centroid position, expressed as: in, This indicates the position of the center of mass at the point of contact between the two feet. Indicates the desired centroid location of the contact point. Denotes the weighted L2 norm. It is a weight vector; The cost function for the task of determining the centroid position at the contact point; the cost function for the task of determining the centroid linear velocity, expressed as: in, The linear momentum representing the center of mass of the exoskeleton. It is the expected centroid velocity. This represents the expected mass center line momentum. This represents the total mass of the exoskeleton robot and the human. Denotes the weighted L2 norm. It is a weight vector; The cost function for the centroid linear velocity task; the cost function for the foot yaw task, expressed as: in, Indicates the current yaw angle of the left / right foot. Indicates the desired yaw angle for the left / right foot; Indicates the current yaw angle In the world coordinate system A unit vector formed on the xy plane such that it is perpendicular to the world coordinate system. The angle between the x-axis and the x-axis is equal to the desired yaw angle. ； Indicates the desired yaw angle In the world coordinate system A unit vector formed on the xy plane such that it is perpendicular to the world coordinate system. of The angle between the axes is equal to the desired yaw angle. ； Indicates and The transpose of a perpendicular unit vector. Indicates perpendicular to The vector, express transpose, Indicates the left foot. Indicates the right foot. Represents the L2 norm; The cost function represents the foot yaw task; the local coordinate system attitude task cost is expressed as: in, This represents the deviation rotation matrix, which is the rotation required to transition from the desired attitude to the actual attitude. This means converting the deviation rotation matrix into a quaternion. Represents a unit quaternion; The cost function for the local coordinate system pose task is represented by the base quaternion derivative regularization cost, expressed as: in, This represents the quaternion rate of change of the base attitude with respect to time; This represents the expected quaternion rate of change of the base attitude with respect to time. The cost function for the base quaternion derivative regularization task is expressed as: in, This represents the total number of contact points on a single foot of the exoskeleton. This represents the total contact force at all contact points; Indicates the first The expected force ratio at each contact point Indicates the first The expected force at each contact point, where j represents the index of all contact points on a single foot of the exoskeleton. Denotes the weighted L2 norm. It is a weight vector; The cost function for force regularization is given by: The cost function for joint regularization is given by: in, Indicates the actual angle of the joint. This indicates the desired angle that the joint will achieve. Indicates joint velocity. Indicates the adjustment parameter. Denotes the weighted L2 norm. It is a weight vector; The cost function for the joint regularization task is represented by the foot swing height task cost, expressed as: in, express The unit vector of the axis. express The unit vector of the axis. express The unit vector of the axis. Indicates contact point of The height of the direction, Indicates the desired foot swing height. Indicates the horizontal velocity at the point of contact. It is the exoskeleton in the control input. The velocity at each contact point; The cost function for the foot swing height task is represented by the contact control regularization task cost, expressed as: in, This indicates the contact point speed in the control input. Indicates the desired velocity at the contact point. This represents the rate of change of contact force in the control input. This represents the expected rate of change of contact force. This indicates the total number of contact points on a single foot of the exoskeleton. 、 Denotes the weighted L2 norm. 、 It is a weight vector; This represents the velocity tracking cost function for the contact control regularization task. This represents the force rate of change tracking cost function for the contact control regularization task.

7. The dynamic walking exoskeleton control system based on human-machine complementary optimization according to claim 6, characterized in that, In the constraint establishment module, the coupled system model includes: dynamic constraints, equality constraints, and inequality constraints; the dynamic constraints include: contact force change rate constraints. in, Indicates contact point The actual rate of change of contact force This indicates the rate of change of contact force in the control input; This means that the formula holds true for all contact points; contact point velocity constraint: in, Indicates contact point The actual speed Indicates the contact point in the control input. speed, This represents the speed transformation matrix, used to adjust the control input based on the contact point height. System momentum change rate constraint: in, This represents the rate of change of the total momentum of the human-exoskeleton system. This indicates the total number of contact points in the system. Represents the identity matrix. Indicates the first The location of each contact point Indicates the location of the system's centroid. This represents the cross product operation. Represents the cross product matrix. Represents the gravitational acceleration vector. Represents the gravity of the system. Represents the zero vector; centroid velocity constraint: in, Indicates the velocity of the system's center of gravity. Represents the linear momentum of the system; base velocity constraint: in, Indicates the base in the world coordinate system The speed in the middle; Indicates the base in the local coordinate system The speed in the middle; Representing the local coordinate system To the world coordinate system Transformation matrix; Base attitude change rate constraint: in, This represents the actual rate of change of attitude of the base. Indicates the actual rate of change of attitude in the control input; joint velocity constraints: in, This represents the actual joint velocity. This represents the joint speed in the control input; the equality constraint is: in, Indicates contact point In the local coordinate system of the foot The position in the middle, Representing the local coordinate system of the foot To the world coordinate system The transformation matrix; the inequality constraints are: in, Indicates a non-negative control boundary; This represents a non-negative control boundary.

8. The dynamic walking exoskeleton control system based on human-machine complementary optimization according to claim 7, characterized in that, In the problem-solving module, the mathematical expression of the model predictive control optimization problem is as follows: in, Represents the state variables of the exoskeleton. Indicates control input, Represents the task weight vector. Represents the rate of change of a state variable. Represents the state transition function. Represents the constraint function. This represents the minimum height of the center of mass above the ground. The state function representing the height of the centroid along the z-axis. The upper and lower bound thresholds representing angular momentum. Represents the angular momentum of the system. express The unit vector of the axis. Indicates the position of the left foot's contact point. Indicates the position of the right foot contact point. This represents the upper and lower bound thresholds of the height difference between the contact points of the left and right feet along the z-axis.

Images