A method and system for stable walking control of a self-balancing lower limb exoskeleton robot

Abstract

This invention relates to the field of robotics, and particularly to a stable walking control method and system for a self-balancing lower limb exoskeleton robot. The control method includes the following steps: S1, establishing a center-of-mass dynamic model (COM) of the self-balancing lower limb exoskeleton robot, and simplifying the COM to a corresponding linear inverted pendulum model; S2, generating the gait of the SBLLE based on the zero-torque point and preview control method, according to the desired CoM and the trajectory of the foothold; S3, introducing CoM compliance through force / torque sensor calculation compensation, establishing a walking controller based on the spring-loaded inverted pendulum model dynamics, and stabilizing the SBLLE during walking through a physical parameter independent feedback control strategy. This control method improves the walking stability of the self-balancing lower limb exoskeleton robot and the robustness of the control strategy to the physical parameters of both the exoskeleton robot and the user.

Description

Technical Field

[0001] This invention relates to the field of robotics, and in particular to a stable walking control method and system for a self-balancing lower limb exoskeleton robot. Background Art

[0002] Lower limb exoskeleton robots can assist paraplegic patients with motor skills and medical rehabilitation. Currently, there are two technological approaches to lower limb exoskeleton robots.

[0003] One type relies on assistive devices such as crutches or carts to maintain dynamic balance and support the patient's walking, as exemplified by products like ReWalk, Ekso, and Cyberdyne. Among these, Wandercraft has developed the Atalante self-balancing lower limb exoskeleton robot with 12 movable joints. Its control strategy employs Hybrid Zero Dynamics (HZD), a strategy based on whole-body dynamics initially used to stabilize humanoid robots and now applied to the efficient and stable walking of self-balancing lower limb exoskeletons. These products prioritize kinematic metrics such as walking speed and user comfort over stability.

[0004] Another type is the self-balancing exoskeleton robot, such as Atalante, REX, and AutoLEE-I, which can achieve self-balancing walking without any aids. The REX exoskeleton, developed by REX Bionics, can achieve self-balancing walking throughout its entire walking cycle without the aid of a cane. Dr. Li and his colleagues designed a highly free-degree-of-freedom self-balancing exoskeleton and proposed an innovative human-in-the-loop control strategy that allows users to walk without canes or other external stabilizing tools by using obstacle energy. This exoskeleton control strategy relies on the coupled dynamics of the exoskeleton and the user, which means that motion stability depends on the physical parameters of both the exoskeleton and the user. These types of products are more difficult to implement in practice due to their hybrid high-dimensional dynamics, redundant joints, hardware limitations, and model uncertainties.

[0005] In summary, a common problem with walking control strategies for lower limb exoskeleton robots is that the stability of the strategy is highly sensitive to the physical parameters of both the robot itself and the user, such as mass, inertia, and height. Summary of the Invention

[0006] This invention provides a stable walking control method and system for a self-balancing lower limb exoskeleton robot, and proposes a feedback control strategy based on a linear inverted pendulum model that is independent of physical parameters to improve the walking stability and performance of AutoLEE-Ⅱ.

[0007] To address the aforementioned technical problems, this invention provides a stable walking control method for a self-balancing lower limb exoskeleton robot, comprising the following steps: S1, establishing a center-of-mass dynamic model of the self-balancing lower limb exoskeleton robot, and simplifying the center-of-mass dynamic model into a corresponding linear inverted pendulum model; S2, generating the gait of the SBLLE based on the zero-torque point and preview control method, according to the desired CoM and the trajectory of the foothold; S3, introducing CoM compliance through force / torque sensor calculation compensation, establishing a walking controller based on the spring-loaded inverted pendulum model dynamic model, and stabilizing the SBLLE during walking through a physical parameter independent feedback control strategy.

[0008] In some exemplary embodiments, S1 includes the following formula for the inverted pendulum:

[0009]

[0010] Among them, [x CoM , z CoM ] T 、[x zmp , z zmp ] T Let m and g represent the position of CoM, the position of ZMP, the mass of the exoskeleton robot, and the acceleration due to gravity, respectively. Assuming the ground is flat and the height of CoM is constant, the inverted pendulum of SBLLE degenerates into a linear inverted pendulum, and the walking of SBLLE is represented as:

[0011]

[0012] Substituting equation (2) into equation (1), the formula for the LIPM dynamics of the SBLLE walking on flat ground is:

[0013]

[0014] In some exemplary embodiments, S2 includes: after setting the trajectory of the foothold, using preview control to find a stable trajectory for CoM and ZMP, and then... As input to LIPM, equation (3) can be rewritten in state-space form as follows:

[0015]

[0016] Then, equation (4) is discretized, and the control cycle time T is expressed as:

[0017]

[0018] in,

[0019]

[0020] Given a foothold and a ZMP trajectory, the following equation can be minimized by representing the inverse problem as an optimization problem:

[0021]

[0022] Where, Δx k =x k -x k-1 and Δu k =u k -u k-1 Q e R and Q x It is a symmetric nonnegative definite matrix.

[0023] In some exemplary embodiments, S2 further includes: previewing N steps in each control cycle, then the solution to equation (6) is expressed as:

[0024]

[0025] The optimal problem from equation (5) to equation (8) is determined by the Riccati equation, given by the following equation:

[0026] P = A T PA-A T PB(R+B T PB) -1 B T PA+Q (8)

[0027] Where, Q = diag(Q y Q x ), K x and K y It can be given by the following formula:

[0028] [K y K x ] = (R + B T PB) -1 B T PA (9)

[0029] Finally, the gain matrix K p Solve using an iterative algorithm.

[0030] In some exemplary embodiments, in S2, the K p The iterative algorithms specifically include,

[0031] Parameter initialization:

[0032] K p (1)=-K y ;

[0033]

[0034] While j <= Ndo;

[0035] K p (j)=(R+B T PB) -1 B T F(j-1);

[0036]

[0037] j = j + 1;

[0038] End while;

[0039] in,

[0040] In some exemplary embodiments, S3 includes: obtaining a measured ZMP via a force / torque sensor having a fixed control time delay T, wherein the ZMP is represented in the complex domain as:

[0041]

[0042] Where, x zmp It's the real ZMP, but... ZMP detected;

[0043] Substituting equation (10) into equation (3) of the center of mass dynamics, we get:

[0044]

[0045]

[0046] According to equation (12), the state equation of SBLLE's motion on the actual trajectory can be rewritten as:

[0047]

[0048] in,

[0049] The required state equation is given by the following equation:

[0050]

[0051] Subtracting equation (13) from equation (14) yields the difference between the actual value and the expected value, expressed as:

[0052]

[0053] In some exemplary embodiments, S3 further includes: rewriting the walking problem as a linear quadratic optimal control problem, wherein the quadratic performance index is expressed as:

[0054]

[0055] Among them, Q θ R is a symmetric semidefinite real matrix. θ It is a symmetric positive definite matrix of real numbers; the feedback control law is expressed as:

[0056] Δu x =-K θ Δθ x (17)

[0057] Discretize equation (15) and use the Discrete Linear Quadratic Regulator (DLQR) method to calculate the gain matrix K. θ By obtaining K θ The formula for the CoM compensation controller is written as:

[0058]

[0059] Among them, K θ[i] Represents K θ The i-th element; then, the compensation amount Δx of CoM. CoM and It can be obtained through numerical iteration, and is expressed as:

[0060]

[0061] In some exemplary embodiments, S3 further includes: in the real-time control of SBLLE, compensating for CoM Δy CoM Add to the desired trajectory of the CoM to adjust the position of the SBLLE's CoM so that the actual ZMP tracks the desired ZMP trajectory, represented as:

[0062]

[0063] in, This is the required CoM; since the CoM equations for the robot on the x-axis and y-axis are independent and have the same formula, the compensation Δy for the y-axis CoM is derived. CoM ,Right now:

[0064]

[0065] in, This is the required CoM.

[0066] This invention also provides a stable walking control system for a self-balancing lower limb exoskeleton robot. The system includes: a model module for establishing a center-of-mass dynamics model of the SBLLE, simplifying the model to a corresponding LIPM; a generation module for generating the SBLLE's gait based on ZMP and a preview control method, according to the desired CoM and the trajectory of the foothold; and a control module for introducing CoM compliance through force / torque sensor compensation calculations, establishing a walking controller based on the SLIP dynamics model, and stabilizing the SBLLE during walking through a physical parameter independent feedback control strategy.

[0067] The technical solution provided by this invention has at least the following advantages:

[0068] This invention provides a stable walking control method and system for a self-balancing lower limb exoskeleton robot. The control method improves the walking stability of the self-balancing lower limb exoskeleton robot and the robustness of the control strategy to the physical parameters of both the exoskeleton robot and the user. Attached Figure Description

[0069] One or more embodiments are illustrated by way of example with reference to the accompanying drawings. These illustrations do not constitute a limitation on the embodiments, and unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0070] Figure 1 For the inverted pendulum model and the linear inverted pendulum model in the center of mass dynamics model;

[0071] Figure 2a The pre-planned trajectories of CoM and ZMP on the x-axis;

[0072] Figure 2b The pre-planned trajectories of CoM and ZMP on the y-axis;

[0073] Figure 3a The simulation model of the self-balancing exoskeleton AutoLEE-II and the kinematic and actuator configuration of the exoskeleton for the solid robot;

[0074] Figure 3b Simulation models of the self-balancing exoskeleton AutoLEE-II and the exoskeleton of solid robots;

[0075] Figure 3c The simulation model of the self-balancing exoskeleton AutoLEE-II and the physical exoskeleton of the solid robot;

[0076] Figure 4 This is a schematic diagram of the first 6 steps of the AutoLEE-II simulation. The numbers in the diagram indicate the robot's walking sequence.

[0077] Figure 5a This represents the distance of ZMP on the x-axis in the simulation.

[0078] Figure 5b This represents the distance of ZMP on the y-axis in the simulation.

[0079] Figure 6a The expected and actual CoM on the x-axis in the simulation;

[0080] Figure 6b The expected and actual CoM on the y-axis in the simulation;

[0081] Figure 6c This represents the corresponding compensation amount on the x-axis for the stabilizer proposed in the simulation.

[0082] Figure 6d This represents the corresponding compensation amount of the stabilizer proposed in the simulation on the y-axis.

[0083] Figure 7 A snapshot of the AutoLEE-II walking experiment shows a female subject walking forward on a flat surface;

[0084] Figure 8a This represents the distance of ZMP on the x-axis during the experiment.

[0085] Figure 8b This represents the distance of ZMP on the y-axis during the experiment.

[0086] Figure 9a The expected and actual CoM values ​​on the x-axis in the experiment;

[0087] Figure 9b The expected and actual CoM on the y-axis in the experiment;

[0088] Figure 9c This represents the corresponding compensation amount of the stabilizer proposed in the experiment on the x-axis.

[0089] Figure 9d This represents the corresponding compensation amount of the stabilizer proposed in the experiment on the y-axis. Detailed Implementation

[0090] As is known from the background art, a common problem with walking control strategies for lower limb exoskeleton robots is that the stability of the strategy is highly sensitive to the physical parameters of both the robot and the user, such as mass, inertia, and height. This invention focuses on achieving stable walking in a self-balancing lower limb exoskeleton robot and improving the robot's robustness to the physical parameters of both itself and the user.

[0091] I. Center of Mass Dynamics

[0092] The self-balancing lower limb exoskeleton (SBLLE) is simplified into a linear inverted pendulum model (LIPM). This simplified model is used to generate the walking trajectory of the SBLLE, and a stable walking feedback controller is designed to maintain the stability of the SBLLE during walking. The motion of the LIPM in the frontal plane is independent of its motion in the transverse plane, and the forms are consistent. It should be noted that this invention only derives the formula for the frontal plane. The formula for the inverted pendulum is as follows:

[0093]

[0094] Among them, [x CoM , z CoM ] T ,[x zmp , z zmp ] T m and g represent the position of CoM, the position of ZMP, the mass of the exoskeleton robot, and the acceleration due to gravity, respectively.

[0095] Assuming the ground is flat and the height of CoM remains constant, the SBLLE inverted pendulum degenerates into a linear inverted pendulum, as follows: Figure 1 As shown. Walking in this situation can be represented as:

[0096]

[0097] Substituting equation (2) into equation (1), the LIPM dynamics of SBLLE walking on flat ground can be given by the following equation:

[0098]

[0099] As can be seen, the LIPM dynamics in equation (3) make the motion characteristics independent of the physical parameters of the SBLLE and the user, which means that changing the user of the SBLLE will not affect the motion of the SBLLE. Due to the uncertainty of the user's physical parameters of the SBLLE, this property is crucial for the SBLLE and makes the control of the SBLLE easier.

[0100] II. Gait Generation Based on ZMP

[0101] When the robot is considered as a LIPM, the ZMP can be obtained by using Equation (3) to give the CoM and the trajectory of the foothold in the SBLLE. To solve the inverse problem, the trajectory of the foothold is first set, and then a stable trajectory is found for CoM and ZMP using preview control. As input to LIPM, equation (3) can be rewritten in state-space form as follows:

[0102]

[0103] Then, equation (4) is discretized, and the control cycle time T is expressed as:

[0104]

[0105] in,

[0106]

[0107] Given the foothold and the trajectory of the ZMP, the following equation can be minimized by representing the inverse problem as an optimal problem:

[0108]

[0109] Where, Δx k =x k -x k-1 and Δu k =u k -u k-1 Q e R and Q x It is a symmetric nonnegative definite matrix.

[0110] Previewing N steps in each control cycle, the solution to equation (6) can be expressed as:

[0111]

[0112] The optimal problem from equation (5) to equation (8) is determined by the Riccati equation, given by the following equation:

[0113] P = A T PA-A T PB(R+B T PB) -1 B T PA+Q (8)

[0114] Where, Q = diag(Q y Q x ), K x and K y It can be given by the following formula:

[0115] [K y K x ] = (R + B T PB) -1 B T PA (9)

[0116] Finally, the gain matrix K pThe following iterative algorithm can be used to solve the problem, as shown in Table 1.

[0117] Table 1: K p Iterative Algorithm

[0118]

[0119] in,

[0120] like Figure 2a and Figure 2b As shown, the ZMP-based gait pattern controller generates six forward steps for SBLLE, with each iteration previewing for 6.4 seconds. The robot's ZMP (red line) tracks the desired trajectory (green line) well and exhibits stability with the CoM trajectory (blue line). The overshoot of ZMP motion on the y-axis is very small and can be ignored.

[0121] III. Independent Control Strategy for Physical Parameters

[0122] The measured ZMP can be obtained using a force / torque sensor with a fixed control time delay T, and can be expressed in the complex domain as:

[0123]

[0124] Where, x zmp It's the real ZMP, but... ZMP detected. In this invention, the delay time of ZMP is T = 0.001s, which is the same as the control time of SBLLE.

[0125] Substituting equation (10) into equation (3) of the center of mass dynamics, we get:

[0126]

[0127]

[0128] According to equation (12), the state equation of SBLLE's motion on the actual trajectory can be rewritten as:

[0129]

[0130] in,

[0131] The required state equation is given by the following equation:

[0132]

[0133] Subtracting equation (13) from equation (14) yields the difference between the actual value and the expected value, expressed as:

[0134]

[0135] When SBLLE walks, the following factors are important for stability:

[0136] 1) The position of CoM must not fluctuate. That is, the acceleration of CoM should be very small.

[0137] 2) The ZMP should be within the supporting polygon (in feet) and should ensure a large stability margin. Therefore, the ZMP should be consistent with the expected value.

[0138] Therefore, the walking problem can be written as a linear quadratic optimal control problem. The quadratic performance index can be expressed as:

[0139]

[0140] Among them, Q θ R is a symmetric semidefinite real matrix. θ It is a symmetric positive definite matrix for real numbers.

[0141] The feedback control law can be expressed as:

[0142] Δu x =-K θ Δθ x (17)

[0143] Discretize equation (15) and calculate the gain matrix K using the Discrete Linear Quadratic Regulation (DLQR) method. θ [40-42]. Directly adjust K θ A more intuitive, simple, and effective DLQR method can be used instead, which only requires adjusting Q. θ and R θ .

[0144] By obtaining K θ The more specific formula for the CoM compensation controller can be written as:

[0145]

[0146] Among them, K θ[i] Represents K θ The i-th element.

[0147] Then, the compensation amount Δx of CoM CoM and It can be obtained through numerical iteration, and is expressed as:

[0148]

[0149] In the real-time control of SBLLE, the compensation Δx of CoM is... CoMAdd to the desired trajectory of the CoM to adjust the position of the SBLLE's CoM so that the actual ZMP tracks the desired ZMP trajectory, represented as:

[0150]

[0151] in, This is the required CoM.

[0152] Because the CoM equations for the robot on the x-axis and y-axis are independent and have the same formula, the compensation Δy for the y-axis CoM can be easily derived in the same way. CoM ,Right now:

[0153]

[0154] in, This is the required CoM.

[0155] When deriving the SLIP dynamics model from Equation (10) to Equation (21) in the CoM model, the physical parameters of the exoskeleton are eliminated. Therefore, the walking controller based on the SLIP dynamics model is independent of the physical parameters of the exoskeleton and the user. That is, the walking strategy is a control strategy with physical parameters independent. The walking control strategy is essentially to introduce the compliance of CoM by calculating compensation through force / torque sensors to stabilize the exoskeleton robot during walking.

[0156] To verify the effectiveness and stability of the proposed feedback control strategy in exoskeleton walking, simulations were performed using the CoppeliaSim software in a simulated physical environment. Figure 3a , Figure 3b and Figure 3c As shown. First, simulations were performed in a dynamic simulator to verify the stability of the strategy. Second, physical experiments were conducted on the AutoLEE-II, a self-balancing exoskeleton with 12 degrees of freedom (DoF), 6 DoF per leg, to verify the effectiveness of the strategy. More details about AutoLEE-II, such as limb length and physical parameters, are also presented.

[0157] Table 2: Values ​​of various parameters in the simulation

[0158]

[0159] I. Simulation

[0160] Since the CoM of the exoskeleton changes with its configuration during walking, the intersection of the exoskeleton's hip joint line and the sagittal plane is designated as the approximate CoM position of the exoskeleton to simplify CoM calculation. Table 2 shows the parameters set in the simulation.

[0161] Figure 4 The simulation demonstrates that the AutoLEE-II successfully and stably walks forward on a flat surface within the CoppeliaSim dynamic software environment. In the simulation, the single-leg support phase and the two-leg support phase were set to account for 80% and 20% of the total walking phase, respectively. Actual ZMP (Zero Footprint Maximum) was calculated based on the foothold, as information about actual footholds was lacking in both the simulation and physical experiments. Figure 5a and Figure 5b As shown, the actual ZMP can track the required ZMP well on the x and y axes and remain within the supporting polygon, which means that the exoskeleton is stable and will not tip over during walking.

[0162] Figure 6a , Figure 6b , Figure 6c and Figure 6d This illustrates the CoM and the compensation responses along the nx and y axes when the exoskeleton moves forward in a straight line. From Figure 6a and 6b A very obvious phenomenon can be observed. Whether on the x-axis or y-axis, the actual CoM of the exoskeleton does not perfectly coincide with the expected CoM position; the actual CoM oscillates around the expected CoM. This phenomenon is due to a discrepancy between the exoskeleton model and the simulation. In trajectory planning, the exoskeleton is treated as a CoM model with concentrated mass, but in actual simulation, the exoskeleton is a robotic system with a continuous physical distribution. Figure 6c and 6d It can be seen that when the exoskeleton foot transitions from the swing phase to the support phase, there are compensating vibrations along the x and y axes, which are caused by the early foot impact.

[0163] II. Experiment

[0164] In the previous section, the effectiveness of the control strategy was verified through simulation scenarios. To evaluate the stability of the control strategy and the mechanical performance of the AutoLEE-II exoskeleton robot, which can provide walking assistance to paraplegic patients, an experiment was conducted on the AutoLEE-II involving a female subject (age 31, weight 48 kg, height xx cm) walking forward in a straight line on a flat surface. Because the control strategy is based on center-of-mass dynamics (in the next section), independent of the physical parameters of the exoskeleton and the main body, the physical parameters were not available beforehand. The experimental surface was not ideally flat and differed from the ideal simulated environment (e.g., unexpected unevenness), causing the exoskeleton's feet to touch the ground prematurely, resulting in a significant impact during actual walking. The experimental results and analysis are shown in the following paragraphs.

[0165] Figure 7 and Figure 8a , Figure 8bThe images show snapshots of AutoLEE-II moving forward on a flat surface and the ZMP response on the x and y axes. Figure 7 As shown, under the proposed feedback control strategy, the exoskeleton robot can walk smoothly and stably alongside the user without any external support, demonstrating that this control strategy achieves self-balancing walking for the exoskeleton robot. Figure 8a and Figure 8b In the model, the ZMP on the x and y axes converges from the desired trajectory. When the exoskeleton's foot touches the ground from the swing phase to the support phase, the ZMP deviates from the expected value and oscillates around it. It eventually converges to the expected value. This is mainly due to differences between the ideal model and the actual exoskeleton model, such as exoskeleton model errors and imperfectly flat ground, leading to a larger impact from premature landing of the exoskeleton's foot on a flat surface. For example, in the single-leg support phase, the linkages and joints are not rigid enough and are flexible, undergoing elastic deformation under gravity, causing the exoskeleton to tilt forward and the swing leg to touch the ground prematurely. Figure 8a , Figure 8b It is easy to see that the maximum absolute errors of ZMP on the x-axis and y-axis are 0.0556m and 0.0460m, respectively. Therefore, the ZMP of the exoskeleton supports the polygon, meaning that the exoskeleton's feet are very stable when walking and do not tip over.

[0166] Figure 9a , Figure 9b , Figure 9c and Figure 9d The response of CoM and x-axis and y-axis compensation is shown when a walking control strategy is employed. From Figure 9a and 9b As shown in the top two figures, the actual CoM converges to the desired CoM, and there is a deviation between the actual and desired CoM. The main reason for this is that the difference between the theoretical model and the actual exoskeleton necessitates stabilizing the robot during walking. The walking control strategy calculates compensations for the x and y axes using equations (20) and (eq: eq21.1), and then superimposes these compensations onto the desired CoM to stabilize the exoskeleton. Figure 9c and 9d As shown, the maximum compensation axes for the x-axis and y-axis are less than 3mm and 6mm, respectively. It should be noted that... Figure 9a , Figure 9b , Figure 9c and Figure 9d The landing impact caused by the transition between the swing leg and the supporting leg cannot be presented. The compliance of the walking control strategy and the flexibility of the mechanical structure absorb those accidental shocks, and the probability of accidents is very small.

[0167] III. Conclusion

[0168] This invention proposes an innovative physical parameter-independent control strategy for a fully driven, self-balancing lower limb exoskeleton to enhance the mobility of patients with dyskinesia due to spinal or lower limb muscle injuries without crutches or other assistive devices. First, the humanoid exoskeleton robot is approximated as a LIPM (Limited Intensity Parallel Model) and physical parameters are eliminated from the mathematical model, resulting in a centroid dynamic equation without any physical parameters. Then, preview control is applied to trajectory planning using CoM (Coordinate Motion) and ZMP (Zero Motion Perspective), and a physical parameter-independent feedback control strategy is proposed to stabilize the exoskeleton during walking. A walking experiment was then conducted on the self-balancing exoskeleton AutoLEE-II with a female user. Through the novel feedback control strategy, AutoLEE-II achieved stable self-balancing walking. Experimental results demonstrate that the feedback control strategy is effective and stable during walking. Future experiments on AutoLEE-II with more diverse subjects will be conducted to verify the robustness of exoskeleton and subject physical parameters to the feedback control strategy.

[0169] Those skilled in the art will understand that the above-described embodiments are specific examples of implementing this application, and in practical applications, various changes in form and detail may be made without departing from the spirit and scope of this application. Any person skilled in the art can make their own modifications and alterations without departing from the spirit and scope of this application; therefore, the scope of protection of this application should be determined by the scope defined in the claims.

Claims

1. A method for stable walking control of a self-balancing lower limb exoskeleton robot, characterized in that, Includes the following steps: S1. Establish a center-of-mass dynamic model of the self-balancing lower limb exoskeleton (SBLLE), and simplify the center-of-mass dynamic model into the corresponding linear inverted pendulum model (LIPM). S2. Based on the Zero Moment Point (ZMP) and preview control method, the gait of the SBLLE is generated according to the desired CoM and the trajectory of the foothold; CoM represents the center of mass; S3. The compliance of CoM is introduced by calculating compensation through force / torque sensors. A walking controller is established based on the dynamic model of Spring Loaded Inverted Pendulum (SLIP). The SBLLE is stabilized during the walking process by the physical parameter independent feedback control strategy. S1 includes, The formula for an inverted pendulum is: Among them, [x CoM , z CoM ] T 、[x zmp , z zmp ] T m and g represent the position of CoM, the position of ZMP, the mass of the exoskeleton robot, and the acceleration due to gravity, respectively. Assuming the ground is flat and the height of CoM remains constant, the inverted pendulum of SBLLE degenerates into a linear inverted pendulum, and the SBLLE walk is represented as: Substituting equation (2) into equation (1), the formula for the LIPM dynamics of the SBLLE walking on flat ground is: S2 includes: After setting the trajectory for the foothold, use the preview control to find a stable trajectory for CoM and ZMP. As input to LIPM, equation (3) can be rewritten in state-space form as follows: Then, equation (4) is discretized, and the control cycle time T is expressed as: Where, x k+1 Let B be the state vector at step k+1; B is the input matrix; u k Represented as an input vector; x k Let k be the state vector at step k. Given a foothold and a ZMP trajectory, the following equation can be minimized by representing the inverse problem as an optimization problem: Where, Δx k =x k -x k-1 and Δu k =u k -u k-1 Q y R and Q x It is a symmetric nonnegative definite matrix; S2 further includes: If N steps are previewed in each control cycle, the solution to equation (6) is expressed as: The optimal problem from equation (5) to equation (8) is determined by the Riccati equation, given by the following equation: P=A T PA-A T PB(R+B T PB) -1 B T PA+Q (8) Where, Q = diag(Q y Q x ), K x and K y It can be given by the following formula: [K y ,K x ]=(R+B T PB) -1 B T PA (9) Finally, the gain matrix K p Solve using an iterative algorithm; K p Iterative algorithms include: Parameter initialization: K p (1)=-K y ; While j <= N do; K p (j)=(R+B T PB) -1 B T F(j-1); j = j + 1; End while; in, 2. The stable walking control method for a self-balancing lower limb exoskeleton robot according to claim 1, characterized in that, S3 includes: The ZMP is measured using a force / torque sensor with a fixed control time delay T, and the ZMP is represented in the complex domain as: Where, x zmp It's the real ZMP, but... ZMP detected; Substituting equation (10) into equation (3) of the center of mass dynamics, we get: According to equation (12), the state equation of SBLLE's motion on the actual trajectory can be rewritten as: in, The required state equation is given by the following equation: Subtracting equation (13) from equation (14) yields the difference between the actual value and the expected value, expressed as:

3. The stable walking control method for a self-balancing lower limb exoskeleton robot according to claim 2, characterized in that, S3 further includes: The walking problem can be written as a linear quadratic optimal control problem, with the quadratic performance index expressed as: Among them, Q θ R is a symmetric semidefinite real matrix. θ It is a symmetric positive definite matrix of real numbers; The feedback control law is expressed as: Thu x =-K θ Dth x (17) Discretize equation (15) and use the Discrete Linear Quadratic Regulator (DLQR) method to calculate the gain matrix K. θ ; By obtaining K θ The formula for the CoM compensation controller is written as: Among them, K θ[i] Representing K θ The i-th element; Then, the compensation amount Δx of CoM CoM and It can be obtained through numerical iteration, and is expressed as:

4. The stable walking control method for a self-balancing lower limb exoskeleton robot according to claim 3, characterized in that, S3 further includes: In the real-time control of SBLLE, the compensation Δx of CoM is... CoM Add to the desired trajectory of the CoM to adjust the position of the SBLLE's CoM so that the actual ZMP tracks the desired ZMP trajectory, represented as: in, It is the required CoM; Since the CoM equations of the robot on the x-axis and y-axis are independent and have the same formula, the compensation Δy for the CoM on the y-axis is derived. CoM ,Right now: in, This is the required CoM.

5. A stable walking control system for a self-balancing lower limb exoskeleton robot, the system being used to implement the stable walking control method for the self-balancing lower limb exoskeleton robot as described in any one of claims 1 to 4, characterized in that, The system includes: The model module is used to establish the centroid dynamic model of SBLLE and simplify the centroid dynamic model into the corresponding LIPM. A generation module is used to generate the gait of the SBLLE based on the desired CoM and foothold trajectory using ZMP and preview control methods. The control module is used to introduce CoM compliance by calculating compensation through force / torque sensors, establish a walking controller based on the SLIP dynamic model, and stabilize the SBLLE during the walking process through a physical parameter independent feedback control strategy.

Images