Lower limb exoskeleton gait strategy optimization method

Abstract

The invention relates to a lower limb exoskeleton gait strategy optimization method. The method comprises a gait strategy multi-objective optimization function and a gait strategy multi-objective optimization method. In order to solve the problems of poor walking stability and high system energy consumption caused by a large amount of human-machine-environment interaction in the walking process of a patient wearing the lower limb exoskeleton, lower limb exoskeleton dynamics modeling and analysis are carried out, and bottom layer gait parameters influencing the walking stability and the system energy consumption are excavated; a gait strategy multi-objective optimization function with ZMP stability margin and single-step driving energy consumption as sub-objectives is established, a gait strategy multi-objective optimization method is provided based on a multi-mechanism optimization strategy cooperative regulation theory, the walking stability of a man-machine system is improved, the energy consumption of the system is reduced, and technical support is provided for improving man-machine coupling of a lower limb exoskeleton system.

Description

Technical Field

[0001] The invention belongs to the field of robotics and rehabilitation medicine, relates to a robot training system, and in particular to a lower limb exoskeleton gait strategy optimization method. Background Art

[0002] Stroke is a disease that causes damage to the brain nerves and surrounding tissues due to sudden rupture or blockage of cerebral blood vessels. The latest data released by the National Health Commission of China shows that stroke has become the leading cause of disability and death among people over 18 years old in my country. The number of stroke patients in my country is increasing at a rate of 9% per year. Currently, the number of patients exceeds 28 million. If no intervention measures are taken, the number of patients will increase to 31.77 million by 2030. At present, there is an extremely unbalanced problem in the demand and supply of rehabilitation medical resources in my country. Lower limb exoskeletons, as an effective auxiliary means, play a positive role in alleviating this social contradiction.

[0003] Foreign countries started research on lower limb exoskeleton robots early and have formed a relatively complete research system. Among them, the three robots Lokomat, ReWalk and HAL are the most representative. Although China started research on lower limb exoskeleton robots relatively late, enterprises and universities represented by Fourier Robotics, Tianjin University and Chengdu University of Electronic Science and Technology have also developed various types of lower limb exoskeleton robots. In the field of lower limb exoskeleton robots, researchers mainly conduct research on the mechanical characteristics and control effects of the robot itself. There is still insufficient research on the development of lower limb exoskeleton robot systems based on the patient's own gait movement characteristics. Therefore, the existing lower limb exoskeleton robots cannot allow patients to fully utilize their own conditions to actively participate in rehabilitation training.

[0004] For lower limb exoskeleton systems that have high requirements for multiple aspects of performance, multi-objective optimization methods are an important technical means to assist them in completing stable and efficient training tasks. Multi-objective optimization methods can find a relatively balanced global optimal solution, which can no longer optimize one or several of the optimization objectives, while not making other optimization objectives worse, and can take into account the improvement of multiple performance indicators. At present, there are few applications of multi-objective optimization methods to lower limb exoskeleton gait optimization.

[0005] Patent CN116502105A discloses an exoskeleton gait parameter optimization method for lower limb rehabilitation training. First, the initial gait of the lower limb rehabilitation exoskeleton is set, and the user wears the lower limb rehabilitation exoskeleton to perform a walking experiment and obtains the exoskeleton-gait fit under the current gait; then, according to the exoskeleton-gait fit corresponding to the gait of the current walking experiment, the gait of the walking experiment in the gait library is clustered using a mixed Gaussian model to obtain a gait clustering cluster, and then the utility of the corresponding gait in the gait library is updated; then, the gait in the gait library is sampled to obtain a sampled gait set, and then the gait with the largest utility is selected from the gait set, and the gait is continuously optimized. Finally, the gait parameters corresponding to the final gait are used as the optimal gait parameters. The present invention can predict the gait parameters of the lower limb exoskeleton that meet the needs of users, and has low calculation cost, fast convergence speed, and high robustness. However, existing technologies do not consider the impact of gait changes on other performance parameters, such as stability and energy consumption of the exoskeleton system, making it difficult to achieve performance balance of the exoskeleton system.

[0006] Patent CN113305812A discloses a bionic lower limb exoskeleton configuration and a method for optimizing the human-machine coupling gait trajectory. The bionic lower limb exoskeleton configuration of the present invention includes three main parts: a bionic hip joint exoskeleton, a bionic knee joint exoskeleton and a bionic ankle joint exoskeleton. The steps of the method for optimizing the human-machine coupling gait trajectory are as follows: 1) a double-chain rigid body simulation model consisting of the coupling of the human physiological structure and the exoskeleton mechanical structure is built in combination with virtual simulation technology; 2) a genetic optimization algorithm is used to conduct a simulation experimental study on the gait trajectory of the human-machine coupling system, so that the gait trajectory of the human-machine coupling system is closer to the natural state. However, the prior art only considers the optimization of the gait trajectory to achieve an anthropomorphic effect, and does not take into account that the lower limb exoskeleton system needs to balance multiple factors to achieve stable and efficient walking.

[0007] Through searching the published patent documents, no published patent documents identical to the present patent application were found. Summary of the invention

[0008] The purpose of the present invention is to supplement the deficiencies of the prior art, improve walking stability, reduce system energy consumption, and enable the exoskeleton robot to more stably and effectively assist stroke patients in rehabilitation training through a lower limb exoskeleton gait strategy optimization method.

[0009] The present invention solves the technical problem by the following technical solutions:

[0010] A method for optimizing gait strategy of a lower limb exoskeleton is provided, wherein the method is characterized by a gait strategy multi-objective optimization function and a gait strategy multi-objective optimization method.

[0011] Prioritize, the construction of the multi-objective optimization function of the gait strategy should be based on the indicators that need to be optimized for the research object. For the lower limb exoskeleton robot, the primary task is to stably and efficiently complete the training movements. Especially for stroke patients, there are differences in the motor ability of the healthy and affected lower limbs. Therefore, when the wearer wears the lower limb exoskeleton for training, higher requirements are placed on the walking stability of the human-machine system. At the same time, the excessive energy consumption of the lower limb exoskeleton system will increase the battery volume and weight, affecting the overall performance of the lower limb exoskeleton.

[0012] Preferably, the multi-objective optimization function of the gait strategy includes two sub-objectives: zero moment point (ZMP) stability margin and single-step drive energy consumption.

[0013] Preferably, the multi-objective optimization method of gait strategy adopts the design concept of a multi-objective optimization algorithm based on the coordination of multi-mechanism optimization strategies, and utilizes the advantages of different optimization algorithms at different optimization stages to improve the optimization effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A schematic diagram of the workflow of a method for optimizing a lower limb exoskeleton gait strategy in an embodiment of the present invention;

[0015] Figure 2 A diagram of a lower limb exoskeleton model in an embodiment of the present invention;

[0016] Figure 3 A schematic diagram of a single-leg model of a lower limb exoskeleton in an embodiment of the present invention;

[0017] Figure 4 Schematic diagram of ZMP ranges in different gait stages in an embodiment of the present invention;

[0018] Figure 5 4 is a flowchart of the gait strategy optimization algorithm in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] For stroke patients, there are differences in the motor abilities of the healthy and affected lower limbs. Therefore, when the wearer wears a lower limb exoskeleton for training, completing the training movements stably and efficiently is the primary task.

[0020] The lower limb exoskeleton is subjected to dynamic modeling analysis. The lower limb exoskeleton model is shown in the figure below. Figure 2 As shown, the single-leg simplified model is as follows Figure 3 According to the generalized force equation of the lower limb exoskeleton, the generalized moment is obtained, as shown in formula (1).

[0021] Among them, θ is the 2×1 order lower limb exoskeleton joint angle matrix, including the hip joint angle θ H and knee angle θ K; M(θ) is a 2×2 order inertia matrix; is a 2×2 order Coriolis force matrix; G(θ) is a 2×1 order gravity matrix; τ is a 2×1 order joint force matrix.

[0022] Through the dynamic modeling and analysis of the lower limb exoskeleton, in order to achieve stable walking, it is necessary to constrain the joint movement angle and speed of the lower limb exoskeleton in different gait stages. In addition, many existing lower limb exoskeleton systems are battery-powered, such as the ReWalk exoskeleton robot. Excessive drive energy consumption will seriously reduce the working time of the system and cannot meet the training needs of patients. If the power output of the motor is reduced, the end position of the lower limb exoskeleton may not reach the target position, affecting the walking stability. Therefore, in order to balance the conflict between walking stability and system energy consumption, this example constructs a gait strategy multi-objective optimization function with ZMP stability margin and single-step drive energy consumption as sub-goals, and proposes a gait strategy multi-objective optimization method based on the coordination of multi-mechanism optimization strategies, which improves the walking stability of the lower limb exoskeleton human-machine system and reduces the system energy consumption, providing technical support for improving the human-machine coupling of the lower limb exoskeleton system.

[0023] The optimization of the lower limb exoskeleton system needs to take into account both walking stability and system energy consumption. The optimization of one goal may reduce the performance of another goal, so this problem belongs to a multi-objective optimization problem.

[0024] Walking stability is one of the two objectives of multi-objective optimization of lower limb exoskeleton systems. In the field of robotics, ZMP is often used as an important stability criterion to evaluate the walking stability of robots. In 1972, Vukobratovic proposed the ZMP criterion for evaluating walking stability, indicating that if the ZMP of a robot always falls within the support domain during walking, the robot will not become unstable. The ZMP range of different gait stages is as follows: Figure 4 shown.

[0025] According to the definition of ZMP, the ZMP trajectory is calculated by the combined force of gravity and inertia. The generalized force matrix F of gravity and inertia is:

[0026] Among them, m i is the mass of the ith link of the lower limb exoskeleton, n is the total number of links of the lower limb exoskeleton, n=5, and are the acceleration components of the ith link in the X, Y and Z directions respectively, and g is the gravitational acceleration.

[0027] The moment matrix of F in each coordinate axis direction is:

[0028] In order to achieve stable walking, it is necessary to make the M of the human-machine system f is equal to 0, so it must satisfy:

[0029] Among them, x zmp and zmp are the positions of ZMP on the X-axis and Y-axis respectively.

[0030] Substituting equation (2) and equation (3) into equation (4), we can obtain:

[0031] In order to quantitatively analyze the stability of the ZMP in the horizontal direction during the robot's motion, this example sets the ZMP stability criterion (J z ), which is mathematically described as:

[0032] Among them, r x (i), r y (i) represents the distance components from ZMP to the foot center of mass on the X-axis and Y-axis respectively, and a z and b z are the weighting coefficients of ZMP along the X-axis and Y-axis, respectively, z +b z =1,a z / b z =L f / W f , L f , W f Represents the length and width of the foot respectively.

[0033] Deflection margin (J b ) is widely used in the field of bipedal robots as a stability criterion in the vertical direction. In the absence of external interference, the lower limb exoskeleton mainly uses the friction torque to offset the influence of the deflection torque on stability, ensuring the torque balance of the lower limb exoskeleton in the vertical direction. Deflection torque J b The mathematical expression is: J b =||M b ||2 Formula (8)

[0034] Among them, M b is the deflection moment. b is defined as:

[0035] Finally, the gait stability objective function, namely the ZMP stability margin, was constructed, and its expression is: f1(x sp )=s z J z +s b J b Formula (10)

[0036] Among them, s z ,s b is a constant, s z ,s b ∈[0, 1], adjusted according to the changes of ZMP in different directions during the walking process of the lower limb exoskeleton; x sp is the set of decision variables.

[0037] In this example, the system energy consumption of the lower limb exoskeleton is analyzed as an optimization target, with the minimum single-step drive energy consumption as the optimization goal.

[0038] The average power method is often used to analyze system energy consumption. The power of the lower limb exoskeleton is equal to the product of the motor output torque and the joint angular velocity. Combined with the lower limb exoskeleton dynamics equation (1), the single-step drive energy consumption (P R ) is:

[0039] Among them, T SW is the gait cycle, M i is the motor driving torque. When i=1, it is the hip joint motor driving torque; when i=2, it is the knee joint motor driving torque. is the joint angular velocity. When i=1, it is the hip joint angular velocity; when i=2, it is the knee joint angular velocity.

[0040] During the walking process of the lower limb exoskeleton, due to the interference between the human, the machine and the environment, the instantaneous power of the motor may increase suddenly, resulting in a decrease in the accuracy of the average power. In addition, the instantaneous power will affect the walking stability of the lower limb exoskeleton system. In order to solve this problem, this example introduces the root mean square deviation distribution of the instantaneous power relative to the average power, χ P , to express the degree of discreteness between the two, the expression is:

[0041] Among them, P I is the instantaneous power.

[0042] Therefore, the single-step driving energy consumption function of the lower limb exoskeleton is obtained as follows: f2(x sp )=P R +χ P Formula (14)

[0043] The multi-objective optimization problem of the lower limb exoskeleton gait includes two optimization objectives: ZMP stability margin and single-step drive energy consumption. This example constructs a multi-objective optimization function F(x sp ), F(x sp )=γ1f1(x sp )+γ2f2(x sp ) Formula (15)

[0044] Among them, γ1, γ2 are constants, γ1, γ2∈[0, 1], which are set according to the influence of ZMP stability margin and single-step driving energy consumption on the performance of the lower limb exoskeleton system during the walking process of the lower limb exoskeleton.

[0042] F(x sp ) is smaller, which means the walking stability of the lower limb exoskeleton is better and the energy consumption of the system is lower.

[0046] The multi-objective optimization method of gait strategy adopts the design idea of ​​a multi-objective optimization algorithm based on the coordination of multi-mechanism optimization strategies, and utilizes the advantages of different optimization algorithms at different optimization stages to improve the optimization effect. The particle swarm (PSO) algorithm has significant advantages in terms of convergence speed, but the PSO algorithm has the problem of being easily trapped in the local optimum. Therefore, this example designs an improved beetle swarm (BSO) algorithm based on the PSO algorithm, so that individuals can effectively jump out of the local optimal solution. The individuals that jump out become new learning objects, which increases the information exchange between individuals and enables the population to find the global optimal solution.

[0047] When using the PSO algorithm to solve the optimal solution of the multi-objective optimization problem of gait strategy, assuming that the population has n particles, it can be obtained that in the D-dimensional space (the number of variables to be solved), the position vector of the kth particle at the tth iteration is The velocity vector is Where k = 1, 2, ... n. Substitute the particle's position vector into the gait multi-objective optimization function F(x sp ), we can use F(x sp ) calculates the fitness function value of the particle, and then judges the quality of the particle by the fitness function value. In each update process, the particle optimizes by tracking two optimal solutions, one of which is the individual optimal solution, denoted as It refers to the position information of the optimal solution found by a single particle in the group; the other is the global optimal solution, recorded as It refers to the position of the optimal solution found by the whole population. Each time the particle swarm evolves, the speed and position of the particles in the swarm are updated according to equations (16) and (17).

[0048] Among them, ω kdis the inertia weight, which represents the inheritance of the particle's velocity from the previous generation; c1 is the individual learning factor, which reflects the influence of the particle's own memory on the particle's motion; c2 is the global learning factor, which reflects the influence of the population's memory on the particle's trajectory; rand() is a random number between [0, 1].

[0049] When the PSO algorithm searches for the global optimal solution in the solution space, the particle group may find a relative optimal solution. The individuals in the particle group will share information so that the population converges to the position of the relative optimal solution. If this relative optimal solution is not the global optimal solution, the algorithm will fall into the problem of local optimal solution. Compared with the PSO algorithm, the BAS algorithm only needs one individual to search for the optimal solution, which greatly reduces the amount of calculation. However, the local search capability of the BAS algorithm is insufficient. It can continuously jump out of local extreme values ​​during the iteration process, which will also lead to insufficient search of these local minimum areas. This example combines the BAS algorithm with the PSO algorithm to design the BSO algorithm, which not only overcomes the problem that the PSO algorithm is prone to fall into the local optimal state, but also uses the group search characteristics of the PSO algorithm to solve the problem of poor local search capability of the BAS algorithm.

[0050] The left and right whiskers of the longhorn beetle can obtain the odor value of nearby food and jump to the side with larger odor value until the global maximum food odor value is found and the global optimal solution is determined. In the D-dimensional space, the center position of the longhorn beetle at the tth iteration is The left and right whiskers of the longicorn are defined as The calculation formula is:

[0051] in, is the distance from the beetle center to the whisker at the tth iteration, expressed as:

[0052] in, is the step size of the tth iteration, and e0 is the relationship parameter between the step size and the distance from the centroid of the longhorn beetle to the whiskers.

[0053] Among them, e1 is the step size attenuation coefficient, and e2 is the step size adjustment parameter.

[0054] is a random unit vector expressed as:

[0055] Wherein, rand(D,1) is a D-dimensional vector consisting of random numbers between [-1, 1], and ||rand(D,1)|| is the norm of rand(D, 1).

[0056] The position of the longicorn will be updated according to the odor concentration detected by the left and right whiskers. The position update formula is:

[0057] in, and are the fitness function values ​​of the left whisker and the right whisker respectively, and sgn() is the sign function.

[0058] The target optimization based on the PSO algorithm mainly relies on group collaboration. The global optimal solution is searched through information exchange between individuals in the group. In the BAS algorithm, the optimization is mainly carried out on an individual basis, without considering the information sharing between individuals in the group. Therefore, this paper combines the PSO algorithm with the BAS algorithm to effectively improve the global search capability of the population. At the same time, the step-size jumping mechanism of BAS is used to solve the problem that the population is prone to fall into the local optimal solution.

[0059] The BSO algorithm optimization flow chart is as follows Figure 5 As shown in , the specific optimization process is as follows:

[0060] Step 1: Set the number of individuals n in the population, set the population dimension and solution space range, and initialize the initial position of the individual. Set parameters: learning factors c1, c2, inertia weight ω kd , the initial jumping step length of the longhorn beetle

[0060] Step 2: In the process of optimizing using the BSO algorithm, the decision variable to be optimized is selected as the individual x sp , according to the gait strategy multi-objective function F(x sp ) Calculate the fitness function value f(x) of each individual at different positions sp ), and then determine the individual optimal solution And the global optimal solution of the population

[0061] Step 3: During the iteration process, the fitness function values ​​of all individuals are compared with the historical maximum fitness function value. If the fitness function value of the current position is greater than the historical maximum fitness function value, the global optimal solution of the population is updated, and the speed and position of the population are updated using equations (16) and (17);

[0062] Step 4: If the individual fitness function value is equal for more than 30 consecutive times, use the position update formula (22) to update the position of each individual;

[0063] Step 5: Calculate the fitness function values ​​of the left and right whiskers of the longhorn beetle to determine the jumping direction of the individual;

[0064] Step 6: Determine whether the algorithm meets the termination conditions. If so, output the optimization result and exit.

[0065] The advantages and beneficial effects of the present invention are:

[0066] 1. The present invention proposes a gait strategy optimization method to balance the conflict between the walking stability of the lower limb exoskeleton system and the system energy consumption, improve the walking stability of the human-machine system, and reduce the system energy consumption.

[0067] 2. The present invention proposes a multi-objective optimization method for gait strategy that coordinates multiple mechanism optimization strategies to solve the problems of slow convergence and easy falling into local optimality in traditional gait optimization algorithms.

Claims

1. A method for optimizing gait strategy of lower limb exoskeleton, characterized in that: The following steps are involved: 1) The dynamic modeling and analysis of the lower limb exoskeleton were carried out, the underlying gait parameters that affect walking stability and system energy consumption were explored, and a multi-objective optimization function of the gait strategy was constructed; 2) The design idea of ​​a multi-objective optimization algorithm based on the coordination of multi-mechanism optimization strategies was adopted, and the advantages of different optimization algorithms in different optimization stages were utilized. A multi-objective optimization method for gait strategy was proposed to balance the conflicting relationship between walking stability and system energy consumption.

2. The lower limb exoskeleton gait strategy optimization method according to claim 1, characterized in that: The multi-objective optimization function of the gait strategy, for the lower limb exoskeleton robot, the primary task is to complete the training movements stably and efficiently. To achieve stable walking, it is necessary to constrain the ZMP of the lower limb exoskeleton in different gait stages. In addition, many existing lower limb exoskeleton systems are battery-powered. Excessive driving energy consumption will seriously reduce the working time of the system and cannot meet the training needs of patients. If the power output of the motor is reduced, the end position of the lower limb exoskeleton may not reach the target position, affecting the walking stability. Therefore, in order to balance the conflicting relationship between walking stability and system energy consumption, a multi-objective optimization function of the gait strategy is constructed with ZMP stability margin and single-step driving energy consumption as sub-goals.

3. The method for optimizing the gait strategy of a lower limb exoskeleton according to claim 1, characterized in that: The described gait strategy multi-objective optimization method is based on the design idea of ​​collaborative multi-mechanism optimization strategy, and designs a gait optimization multi-objective algorithm (BSO) to overcome the common problems of slow convergence and easy falling into local optimality in traditional lower limb exoskeleton robot gait optimization schemes. BSO integrates the Tianniu algorithm on the basis of the traditional particle optimization algorithm model, giving full play to the advantages of multiple optimization methods in different optimization stages.

Images