CN117697717A - Exoskeleton physical man-machine two-way interaction simulation system - Google Patents

Abstract

The invention discloses an exoskeleton physical human-machine two-way interactive simulation system, which belongs to the field of exoskeleton robots. The invention includes an exoskeleton human-machine system, a data transmission module, a simulation synchronization module, an intelligent algorithm module, and a simulation visualization module. Different from traditional pure simulation systems, the simulation system proposed by the present invention can be connected to the exoskeleton robot through a data interface to realize real-time reconstruction of the actual system of the exoskeleton robot and its surrounding environment in the simulation environment, and in the simulation environment The model can perform two-way synchronization with the real exoskeleton robot, realizing a physical human-machine two-way interactive simulation system with high real-time feedback.

Description

An exoskeleton physical human-computer two-way interactive simulation system

Technical field

The invention belongs to the field of intelligent wearable devices, especially exoskeleton systems.

Background technique

Exoskeleton robot technology is a typical multi-disciplinary technology, involving professional knowledge from multiple disciplines such as clinical medicine, rehabilitation medicine, mechanical design, circuit systems, information processing, mobile computing, and control algorithms. According to the application objects, it can be divided into power-assisted exoskeleton robots for healthy people's ability enhancement and medical exoskeleton robots for medical rehabilitation training. Medical exoskeleton robots can be divided into two categories: one is rehabilitation training robots for hospital scenarios, and the other is walking assistance exoskeleton robots for home and personal use.

As a typical wearable intelligent system with strong human-machine coupling, exoskeleton robots have a real-time interaction process with the wearer. Due to the large individual differences between the wearers, in order to obtain a more comfortable and natural walking process, the exoskeleton robot needs to have two-way interaction and mutual adaptation with the wearer. Therefore, it requires technologies such as perception recognition, cognitive decision-making, motion planning, and human-computer interaction. There are higher requirements.

As a typical wearable intelligent system that moves in a real environment, exoskeleton robots have real-time interaction processes with the surrounding environment. Since the environment faced by the exoskeleton human-machine system is dynamically changing in real time, and there are some unknown situations at all times, there are high requirements for the exoskeleton robot's perception and recognition of the terrain environment, cognitive decision-making, motion planning, dynamic adaptation and other technologies. .

Since the development of exoskeleton robots requires the cooperation of technical talents from multiple disciplines such as mechanics, computers, embedded, artificial intelligence, and control, its development is difficult and the development cycle is often long. And once the design is completed, testing and iterative updates of its functions also require a large amount of data and result verification, and the actual clinical testing process is often complex and faces many difficulties. Therefore, it is of great significance to build a simulation test system for exoskeleton robots, collect and store the structural model, wearer model and environmental information of exoskeleton robots in real time, and build a two-way interactive simulation system for exoskeleton robots based on this information. , by recording and simulating the test process and applying various perception, decision-making and control algorithms, the development efficiency of exoskeleton robots can be improved, the development cycle can be shortened, and testing efficiency can be improved.

The Chinese patent "ROS-based simulation control platform and exoskeleton robot simulation control system" with publication number CN113608451A discloses a ROS-based simulation control platform and exoskeleton robot simulation control system, which can interact with real exoskeleton robots and interact with data. Perform simulation control. However, this system only provides data interaction and control algorithms between exoskeleton man-machine and ignores the impact of environmental information on the exoskeleton man-machine system. Its application scope is relatively limited.

The Chinese patent "An exoskeleton robot platform communication protocol and online simulation control system" with publication number CN112631148A discloses a communication protocol for the exoskeleton robot platform. Based on this protocol, a communication protocol between the exoskeleton robot client and the remote server can be constructed. The communication process is used to transmit sensing signals and control instructions of the exoskeleton robot. However, the patent does not specify the specific operation process of the simulation system.

The Chinese patent "Lower Limb Movement Simulation Control Equipment for Virtual Reality Environment" with publication number CN105892626A discloses a lower limb exoskeleton robot training system that can provide richer training for the rehabilitation training process by interacting with the virtual reality environment. content to enhance patients’ active participation. However, this patent only provides the process of information interaction and simulation interaction for the virtual environment constructed by the computer, and does not make further design of the real environment faced by the exoskeleton robot in the simulation environment.

To sum up, the current exoskeleton robot simulation system mainly has the following problems:

1. The simulation model in the simulation environment is only for the exoskeleton robot system, and it is not enough to model the status information of the wearer itself;

2. Most of the environmental information such as terrain in the simulation environment is a virtual constructed environment, and there is no modeling of the real environmental information faced by the exoskeleton robot;

3. The interaction method between the simulation environment and the exoskeleton human-machine system is single, only focusing on the data transmission of sensing and control instructions. The synchronization between the simulation model and the real exoskeleton human-machine system is poor.

Contents of the invention

The present invention solves the problem that in the prior art, the simulation model in the simulation environment is only for the exoskeleton robot system, and is not enough to model the state information of the wearer itself; most of the environmental information such as terrain in the simulation environment is a virtual constructed environment, which is not specific to the exoskeleton robot system. The real environment information faced by the exoskeleton robot is modeled; the interaction method between the simulation environment and the exoskeleton human-machine system is single, only focusing on the data transmission of sensing and control instructions. The interaction between the simulation model and the real exoskeleton human-machine system Technical issues with poor synchronization.

In view of the technical problems that need to be solved, the technical solution of the present invention is an exoskeleton physical human-machine two-way interactive simulation system. The system includes: an exoskeleton human-machine system, a data acquisition module, a simulation synchronization module, an intelligent algorithm module, and a simulation visualization module;

The exoskeleton human-machine system includes: an exoskeleton robot, the exoskeleton robot is worn by the wearer, and the skeleton of the exoskeleton robot imitates the limbs below the waist of the human body;

The data acquisition module includes: an EEG signal acquisition module, an IMU sensor, a joint angle sensor, an interactive force sensor, a pressure sensor, a depth camera, and an IMU attitude sensor. The EEG signal acquisition module is worn on the human head; the IMU Sensors, joint angle sensors, interactive force sensors, pressure sensors, and depth cameras are all installed on the exoskeleton robot. The IMU attitude sensor is installed on the wearer of the exoskeleton robot. The IMU sensor detects the attitude of the exoskeleton, so The interactive force sensor detects the force exerted by the wearer on the exoskeleton, the pressure sensor detects the pressure on the soles of the exoskeleton, the IMU posture sensor detects the posture of each part of the human body, and the depth camera collects the road conditions in front of the wearer;

The simulation synchronization module establishes a data exchange channel between the exoskeleton human-machine system in the real environment and the simulation model in the virtual environment, and establishes communication between the simulation environment and the physical human-machine system through an Ethernet-based network communication protocol. protocol and transfer data between the two;

The intelligent algorithm module includes: perception and recognition module, cognition and decision-making module, planning and control module; wherein the perception and recognition module is used to identify the terrain environment, human-machine motion status, etc. from the acquired sensor data. and analysis; the cognition and decision-making module is used to make decisions on subsequent motion paths based on the current environment and motion state of the human-machine system; the planning and control module is used to plan the motion mode of the human-machine system based on the aforementioned cognitive decision-making results And control the human-machine system to complete the movement;

The simulation visualization module includes: a data display module, an algorithm simulation visualization module, and a simulation process playback module; the data display module displays all data from the exoskeleton physical human-machine system and intermediate state data generated during the simulation process; algorithm simulation visualization The module displays the simulation results on the future motion status of the exoskeleton human-machine system in the simulation environment, so that R&D personnel can clearly observe the corresponding algorithm perception, recognition, planning and control results; the simulation process playback module displays all completed history The simulation results and the movement process of the real exoskeleton physical human-machine system are played back, so that developers can observe and compare the state differences of the exoskeleton human-machine system under the action of various algorithms.

Further, the specific calculation steps of the calculation method in the perception and recognition module are:

The perception and recognition method consists of a center of mass estimator, a slope estimator and a Hill muscle model; the center of mass estimator obtains the motion state and stability of the human-machine system through changes in its center of mass when it moves; the slope estimator estimates the slope in the external environment, Use this for gait planning on external ramps;

First of all, in this exoskeleton robot, the ankle joint is set as a passive joint that can move within a small range in order to better assist walking. Therefore, there are five active joints, namely 2 knee joints, 2 hip joints, and pelvic joints. joints; while the wearer considers the joint angles of 10 active motion joints, namely 2 knee joints, 2 hip joints, 2 shoulder joints, 2 elbow joints, lumbar joints, and pelvic joints; each joint is set There is a local coordinate system, and the world coordinate system is set outside the human-machine system;

First, calculate the center of mass position of the human-machine system through the collected data; the formula for determining the center of mass of the human-machine system is as follows:

Among them, C _machine is the center of mass position of the human-machine system, is the displacement vector from the human pelvic joint to the origin of the world coordinate system, M _h and M _e represent the masses of the wearer and the exoskeleton robot respectively, M is the total mass of the human-machine system, C _h and C _e represent the wearer and the exoskeleton robot respectively. The center of mass position of the exoskeleton robot, ε is the designed physical coupling model; where the displacement vector /> It is obtained from the data obtained from the IMU attitude sensor on the wearer; the formula for determining the position of the wearer's center of mass is as follows:

H＝[H ₁ H ₂ …H ₁₀ ] is the wearer’s transfer matrix, where And so on;/> is the uniform transition matrix between adjacent joints,/> with/> are the rotation matrix and displacement vector from joint i to joint j respectively, where the rotation matrix is obtained by the joint angle measured by the joint angle sensor on the wearer, and the displacement vector is obtained by the attitude sensor on the wearer;/> are the kinematics and static parameters of the human body, which are obtained by collecting the static postures of 30 wearers and using formula (2);

The formula for calculating the center of mass position of the exoskeleton robot is as follows:

E＝[E ₁ E ₂ ... E ₅ ] is its state matrix, where And so on;/> are its kinematic and static parameters; since the exoskeleton robot only considers five active motion joints, its transfer matrix and kinematic and static parameters are fifth-order matrices; the formula for calculating the physical coupling model ε is as follows:

ε＝R _Y (θ _lh )δ ₁ +R _Y (θ _lk )δ ₂ +R _Y (θ _rh )δ ₃ +R _Y (θ _rk )δ ₄ , (4)

Among them, θ _lh , θ _lk , θ _rh , and θ _rk are the joint angles of the left hip, left knee, right hip, and right knee respectively. R _Y (θ _i ) represents the rotation matrix of joint i relative to the y axis. Through the joint angle The angle data of the sensor is obtained; δ _j is the weight, j = 1, 2, 3, 4, which is obtained through experiments from the data of the exoskeleton foot pressure sensor;

After calculating the center of mass position of the human-machine system through the above steps, the exoskeleton human-machine system is simplified into a two-dimensional linear inverted pendulum model, and the potential energy of the inverted pendulum model during the motion is calculated through formula (5);

Among them, x and represent the position and velocity information of the center of mass in the horizontal direction respectively, z _c represents the height of the center of mass, and g represents the gravitational acceleration constant;

Orbital energy is used to represent the motion state of the center of mass. During a single swing of the inverted pendulum, the orbital energy remains at a constant value until the next time the supporting legs of the inverted pendulum are switched, and the orbital energy will be recalculated; when the human-machine system moves The orbital energy E _current of the current state is calculated by formula (5); different orbital energies result in different next landing points (capture points) on the horizontal ground, so there are two situations here. When E _current > 0, It means that after completing the next switching of the supporting legs, the inverted pendulum model can continue to move forward; when E _current <0, it means that the inverted pendulum model does not have enough energy to cross the highest point of potential energy and continue to move forward. Before reaching the highest point of potential energy, The speed drops to zero and begins to move backward; therefore, the motion state of the human-machine system can be judged based on the orbital energy;

The slope estimator is used to estimate the slope of the slope where the current exoskeleton human-machine system is located in real time. It is built based on the sensing data and geometric size information of the exoskeleton itself; the IMU sensor located in the exoskeleton backpack can sense the slope of the entire torso in real time. Posture information, angle sensors located in the hip, knee, and ankle joints can sense the angle information of each joint in real time;

Since during normal walking, a person's legs step forward alternately, when one leg acts as a supporting leg, the other leg acts as a swinging leg. Here, it is assumed that the sole of the supporting leg is always in contact with the slope surface and does not slide relative to it. Therefore, the slope estimator calculates the slope based on the joint angles of the hip, knee, and ankle of the supporting leg and the posture information of the trunk. And the left and right legs alternately support each other; as shown in Figure 1, the exoskeleton human-machine system is divided into (a), (b), and (c) according to the different gait phases of the supporting leg and the swing leg when walking on the slope. ), (d) four situations, among which: (a) indicates the moment when the swing leg just leaves the ground during the uphill process; (b) indicates the moment when the swing leg is about to touch the ground during the uphill process; (c) Indicates the moment when the swing leg just leaves the ground during the downhill process; (d) indicates the moment when the swing leg is about to touch the ground during the downhill process; Combining these four situations, we get that during the uphill and downhill walking process, when All joint angles and posture information of the trunk when one leg is used as a supporting leg and the other leg is used as a swing leg and moves forward in the air, and the slope of the current ramp is calculated based on this information and the geometric size information of the exoskeleton human-machine system; Assume that θ represents the slope of the ramp, and the θ angle is defined to be in a similar coordinate system to the joint angle of the exoskeleton, which also follows the right-hand rule, that is, θ<0 during the uphill process, and θ> during the downhill process. 0, θ=0 on the horizontal ground; where α ₁ , α ₂ , α ₃ , and α ₄ respectively represent the interior angles corresponding to the vertices A, B, C, and E in the polygon ABCEFA. Point A represents the location of the hip joint, and point B Represents the position of the knee joint, point C represents the position of the ankle joint, point E represents the intersection of the extension line of the calf link CD and the X-axis, and F represents the coordinate origin of the coordinate system X-0-Y. There is the following geometric relationship:

Based on the plane geometric relationship, the sum of the interior angles of a polygon with N sides is (N-2)·π. For the multi-deformation with 5 sides composed of A-B-C-E-F-A, we get:

Based on formula (6), the calculation results of slope θ in four cases are:

The slope of the ramp is calculated based on the posture information of the trunk and the joint angles of the supporting legs;

The muscle model collects the correspondence between the electromyographic signals, joint angles, etc. and muscle force of healthy volunteers of different body types during rehabilitation training, and introduces the neural network model to build a personalized joint torque prediction model to provide lower limb assistance for different wearers. Provide theoretical support for personalized muscle strength prediction when robots perform rehabilitation training. It abstracts the muscle into a muscle structural mechanics model composed of three elements: series elastic unit (SE), active contraction unit (CE) and parallel elastic unit (PE), and reflects the muscle contraction characteristics. The muscle model is shown in Figure 2. The muscle model is shown in formula (9):

L ^MT =L ^T +L ^M cos a. (9)

Among them, L ^MT is the total length of the tendon and muscle fiber, L ^T is the length of the tendon, L ^M is the total length of the muscle fiber, α is the pennation angle, which is the angle between the muscle fiber bundle and the tendon; the active contraction unit and the parallel elastic unit pass through the feather The shape angle α is connected with the series elastic unit to form a muscle system; the active contraction unit is the active part during the muscle work process, and the parallel and elastic units are the passive contraction part. During the active contraction process, the muscle fibers are attached to the bone surface through stretching. Tendons, thereby driving joint movement; muscle fiber force is the sum of active units and passive units, that is, it is considered that at any time, the total force generated by muscle fibers is F, then:

F＝F _CE +F _PE , (10)

In the formula, the force generated by the F _CE active contraction unit is the active force, and the force generated by the F _PE parallel contraction unit is the passive force; in the muscle model, F _CE represents the force generated by the chemical reaction inside the muscle cross bridge, which can be considered as Produced by the active contraction of muscle fibers, its size is related to the degree of muscle activation, muscle length, and muscle contraction rate; the specific expression of F _CE is:

F _CE =a·F ₀ ·f(l)·f(v), (11)

In the formula, a is the degree of muscle activation, which is related to the amplitude of the electromyographic signal. If the activation degree of a certain muscle at a certain moment is 0, it means that there is no active muscle force in the current muscle, then F _CE =0. The higher the degree of activation, the more muscle fibers are involved in exercise contraction. When muscle activation a=1, it is considered that all muscle fibers in the target muscle are involved in exercise. At this time, the tension generated by the muscle reaches the maximum value; F ₀ is the maximum contraction force of the muscle. , its size is affected by the cross-sectional area of the muscle's physiological shape:

F ₀ =K·PCSA, (12)

In the formula, PCSA is the cross-sectional area of the muscle, and K is the maximum stress that the muscle can provide. The length of muscle fibers and the rate of muscle contraction are related to the angle of the moving joint;

The force generated by the F _PE parallel contraction unit is the passive force, which is generated by the elastic deformation of muscle fibers. When the length of the muscle fiber exceeds a certain range, the muscle will generate tension to force the muscle fiber to return to its original state. Therefore, the size of F _PE is mainly affected by the maximum contraction force (F ₀ ) dynamic length relationship:

F _PE =F ₀ ·f(l). (13)

After calculating the joint muscle strength, by comparing it with the muscle strength data of healthy volunteers and the wearer's muscle strength data, the physiological state of the wearer at this time can be obtained; in active training, it can also be used as output Extract and identify the wearer's active movement intention.

Further, the specific calculation steps of the calculation method in the cognition and decision-making module are:

After obtaining the centroid position and ramp slope based on the perception and recognition algorithm, the calculated data is used to calculate cognitive and decision-making methods;

In order to maintain balance, the orbital energy of the robot is usually set to 0. It is assumed that the exoskeleton robot is currently in a standing balance state. If the exoskeleton robot is disturbed by the outside world (for example, someone applies a push from behind), in order To maintain balance, the exoskeleton robot needs to take a forward step to restore balance. The capture point theory predicts the next foothold based on the current motion state of the center of mass. The exoskeleton robot maintains walking by controlling the foothold of the exoskeleton robot's next step. stability and balance in the process;

When the environment is flat, it is assumed that the process of exchanging supporting legs of the inverted pendulum model is completed instantaneously without losing energy. The next step is calculated based on the center of mass position, center of mass velocity and given orbital energy of the current inverted pendulum model; Assume The orbit energy of the inverted pendulum model after switching the supporting legs is E _desired , then:

where x _t and Respectively represent the position and acceleration of the center of mass in the horizontal direction at the next time the support leg is switched, x _cp represents the next foothold (capture point) on the horizontal ground, and z _c represents the height of the center of mass;

Therefore, given the desired orbital energy E _desired , the orbital energy adaptive capture point x _cp is calculated as:

Among them, the sign() function is a sign function, when The return value is 1, otherwise the return value is -1; the switching time of the supporting leg should be a tested experience value, and the selected switching time must ensure The above formula is meaningful, otherwise the orbital energy cannot reach the _desired Edesired in the next step, and multiple steps are needed; given a desired orbital energy _Edesired , each step is calculated through the current motion state of the center of mass. The foothold of one step, and even if the exoskeleton human-machine system model is interfered by external forces such as crutches during this process, it will ultimately affect the motion state of the center of mass in the current step. When calculating the foothold of the next step, it will be based on the new Calculate the motion state;

When there is a slope in the environment, the height of the linear inverted pendulum model is variable and related to the slope; during the movement of the two-dimensional linear inverted pendulum on the slope, the height of the center of mass moves along a straight line. The slope of is related to the slope of the ramp. The equation of the straight line is:

z＝z ₀ +kx, (16)

Among them, the slope k=tanθ, θ is the slope of the ramp, z ₀ represents the height of the center of mass at the initial moment; it is worth noting that during the uphill movement, θ<0, so k<0; during the downhill movement, θ>0, so k>0; First, let’s look at the motion in the horizontal direction. Whether on flat ground or on a slope, the horizontal motion process of the linear inverted pendulum model is the same. Therefore, as long as we know the motion of the inverted pendulum model at the initial moment The motion state of the center of mass can be calculated as the motion state of the center of mass at any subsequent moment. The method for calculating the motion state of the center of mass is:

in is a constant, t represents time, so that the horizontal motion state of the center of mass at any moment in a single swing cycle of the inverted pendulum can be calculated; in addition, the motion in the vertical direction is obtained by the straight-line formula of the center of mass moving along the ramp, The calculation method is as follows:

z _t =z ₀ +k·Δx＝z ₀ +k·(x _t -x ₀ ) (18)

Among them, z ₀ represents the height of the center of mass at the initial moment, and Δx represents the horizontal movement distance of the center of mass from the initial moment to the moment when the supporting leg switches; then, assuming that the inverted pendulum model is in a certain swing process, the position of the swinging legs at the initial moment is _P0 :

Then when the supporting leg is switched at time t, the foothold of the swing leg is calculated using formula (20):

The starting position P ₀ of the swinging foot and the next landing position P _cp have been obtained. Based on these two positions, a spatial trajectory from the starting position of the swinging foot to the landing point is planned as the gait trajectory of the exoskeleton robot.

Further, the specific calculation steps of the planning and control method are:

After obtaining the starting position and next landing position of the exoskeleton robot's swinging feet based on the cognitive and decision-making algorithm, a gait trajectory of the exoskeleton robot can be planned based on the positions of these two landing points. Since human gait has certain movement patterns, if conventional interpolation methods are used to randomly plan a trajectory, the walking process may not conform to the walking patterns, making the walking process uncomfortable. Therefore, it can be considered to learn the motion rules of gait from the gait trajectories of healthy people and generate new gait trajectories according to requirements.

In this algorithm, the motion capture system Vicon is used to collect the walking trajectory of normal people on flat ground as the teaching trajectory of the gait model. After learning the motion rules of the gait trajectory, the gait model based on dynamic motion primitives can plan different gait trajectories based on the footholds predicted in real time by the exoskeleton human-machine system model, thereby adapting to slope terrain of different slopes. The basic model for modeling the gait trajectory of a lower-limb walking exoskeleton robot is:

where τ is a constant, used to scale the amount of time to adjust the speed of the trajectory, α _y , β _y are both constants, y=[xz] ^T represents the spatial position of the swinging foot, represents its speed and acceleration, y _cp represents the final expected foothold; f is a nonlinear term, as shown below:

N represents the number of basis functions, v comes from a first-order system, and satisfies where a _v is a constant; the basis function Ψ _i is the radial basis function:

Among them, σ _i and c _i respectively represent the width and center value of the basis function Ψ _i ; assuming that the teaching trajectory is expressed as Then we get the f _target that needs to be learned:

To determine the weight value w _i corresponding to each basis function Ψ _i , construct a loss function:

Among them, t represents the time step, and T represents the total number of time steps of the entire trajectory; the minimization process of solving the above loss function is to obtain the parameters of the model. After the gait trajectory model learning is completed, different initial positions and end positions of the trajectory are given. to generate different gait trajectories.

Combining the data calculated by the previous two algorithms, the exoskeleton human-machine system model can be used to predict the foothold of each step on the slopes of different slopes. Based on the known starting point and foothold of each step, combined with the DMP-based The gait trajectory model can generate the gait trajectory and subsequent motion sequence from the starting point of each step to the foothold of the next step.

Compared with traditional robot simulation systems, the simulation system provided by the present invention has the following advantages:

1. The state of the exoskeleton human-machine system between the real world and the simulation environment is better synchronized, and the physical human-machine system of the exoskeleton can be fully digitally reproduced, rather than simply between the exoskeleton human-machine system and the simulation platform. carry out data transmission;

2. The algorithm simulation is richer. Based on the above-mentioned digitally reproduced exoskeleton human-machine system model, the simulation system can use the computing power of the cloud server to perform parallel simulations, and can simultaneously simulate multiple algorithms and perform comparative analysis;

3. The simulation platform has a higher degree of visualization. Unlike traditional simulation systems that can only simply display the data curves during the simulation process, the simulation platform provided by the present invention can synchronously reproduce the motion status information of the exoskeleton human-machine system and display it on the simulation model. display. In addition, the multi-algorithm parallel simulation process and simulation playback function provided by the present invention can also provide richer simulation process visualization results.

Description of the drawings

Figure 1 is a schematic diagram of the four situations of the exoskeleton human-machine system on the ramp;

Figure 2 is a schematic diagram of Hill’s theoretical model;

Figure 3 is a schematic diagram of the architecture of the physical human-machine two-way interactive simulation system;

Figure 4 is a schematic diagram of the sensing system of the exoskeleton robot;

Figure 5 is a schematic diagram of the wearer data collection system;

Figure 6 is a schematic diagram of the environmental data collection system;

Figure 7 is a schematic diagram of the data acquisition module;

Figure 8 is a schematic diagram of the simulation synchronization module;

Figure 9 is a schematic diagram of the intelligent algorithm module;

Figure 10 is a schematic diagram of the simulation visualization module.

Implementation

The invention discloses a physical human-machine two-way interactive simulation system of an exoskeleton robot, which includes an exoskeleton human-machine system, a data transmission module, a simulation synchronization module, an intelligent algorithm module, and a simulation visualization module. As shown in Figure 3.

The present invention is an expansion and improvement of the traditional robot simulation system and needs to rely on the existing operating system. For example, the main control software of the exoskeleton human-machine system can be built based on the embedded Linux operating system, the intelligent algorithm module of the simulation system can be built based on the server/cloud server operating system of Linux, Unix, and Windows, and the visualization module of the simulation system can be built based on Linux, Unix, and Windows server/cloud server operating systems. Construction of desktop operating systems such as Windows and MacOS.

The application objects of the present invention are exoskeleton robot systems and simulation systems based on personal laptops/desktop computers/servers, and each module needs to exchange data through corresponding communication protocols. On the exoskeleton robot system, data transmission is completed between the exoskeleton main control module and other sub-modules of the system through CAN, SPI, I2C, serial port, Ethernet and other communication protocols; the intelligent algorithm module in the exoskeleton robot system and simulation system The communication with the visualization module can be accomplished through advanced communication protocols based on Ethernet, such as TCP/IP, Socket, etc. You can also build your own communication protocol based on Ethernet.

Exoskeleton human-machine system: This part is mainly composed of the actual exoskeleton robot and the wearer in the real world. The exoskeleton robot can interact with the wearer and the real physical environment, and generate various sensing and interaction data in real time, mainly including Three major categories: exoskeleton data, wearer data, and environmental data. Among them, the exoskeleton data includes the movement posture, joint angles, human-computer interaction force, sole pressure and other information of the exoskeleton human-machine system (as shown in Figure 4); the wearer data includes the wearer's own physiological functions such as myoelectricity/encephalography. Signal data (as shown in Figure 5); environmental data includes terrain data and surrounding obstacle data obtained based on attitude sensing and visual sensing (depth cameras, point cloud cameras and other devices) (as shown in Figure 6).

Data collection module: As shown in Figure 7, this module can collect various data of the aforementioned exoskeleton human-machine system in real time (Figure 4, Figure 5, Figure 6), including exoskeleton data, wearer data and environmental data. After these data are converted into digital signals through the digital-to-analog conversion interface (AD conversion module), the sensing data can be transmitted through CAN bus, serial port bus, SPI bus, I2C bus and Ethernet according to the bus protocol used by the sensor module. In the data acquisition module, the data acquisition module then packages all the sensor data into a unified format and transmits it back to the simulation system in real time through the simulation synchronization module. The posture data of the data acquisition module mainly comes from the IMU-based posture sensing system. The module for posture collection in the exoskeleton robot system requires multiple IMU-based posture sensors to be installed on the wearer's head, torso, and arms. At the same time It is also necessary to install IMU-based attitude sensors on the torso, thighs, calves, feet and other parts of the skeletal robot to collect all-round attitude information of the exoskeleton human-machine system. The posture data can be transmitted to the simulation environment through the simulation synchronization module and used for motion state reconstruction and real-time visualization of the exoskeleton human-machine system.

The interactive force data in the data acquisition module mainly comes from 3-axis/6-axis interactive force sensors installed at the thighs, thumbnails, ankles, soles, etc. of the exoskeleton. They are generally installed at the parts where the exoskeleton robot comes into contact with the wearer and are tied to the exoskeleton. Linked together, they are used to collect the interaction force between the exoskeleton robot and the wearer during the entire movement, as well as the contact force between the exoskeleton human-machine system and the ground. The original interaction force data is generally an analog signal, which also needs to be converted into a digital signal through an analog-to-digital conversion interface (AD), and then the data is transmitted to the outside through the corresponding data transmission bus (CAN, SPI, I2C, serial port, etc.) The main control part of the skeletal human-machine system is then transmitted to the simulation environment through the simulation synchronization module.

The wearer's data in the data acquisition module mainly collects the wearer's physiological signals in real time through electromyography (EMG) and electroencephalography (EEG) acquisition equipment. After simple filtering, it is transmitted to the simulation environment through the simulation synchronization module for visualization and processing. The perception recognition module used in intelligent algorithms can be used to characterize the wearer's current motion status, physiological status and other basic information. In addition, this information can also be used to perceive and classify the wearer's movement intentions, trigger movement patterns, predict movement trends, etc.

Environmental data collection in the data collection module is generally completed through visual sensing systems installed on the front, side and rear of the torso of the exoskeleton human-machine system, including depth cameras, RGB cameras, lidar, point cloud cameras and other equipment. The environmental information around the exoskeleton human-machine system is collected in real time, and the current visual sensing posture information is collected in real time through the posture sensor connected to the visual sensing device to identify and correct the current visual sensing data. The visual sensing data and corresponding attitude data are synchronously packaged and transmitted to the simulation environment for reconstruction and visualization of the terrain environment. In addition, the visual sensing data needs to be synchronized to the simulation visualization module and displayed in real time to observe the environment currently faced by the exoskeleton human-machine system from a first-person perspective.

Simulation synchronization module: As shown in Figure 8, the main function of this module is to establish a data exchange channel between the exoskeleton human-machine system in the real environment and the simulation model in the virtual environment, and establish simulation through the Ethernet-based network communication protocol Communication protocol between the environment and the physical human-machine system and transfer data between the two. This module should have a high real-time communication rate and can transmit various sensing data and motion status data of the aforementioned exoskeleton human-machine system to the simulation environment in real time, so that the exoskeleton in the real world can be "reproduced" in the simulation environment. The human-machine model and its surrounding environmental information construct a simulation environment based on real data for subsequent algorithm modules. In addition, the simulation synchronization module can also synchronously send the control instructions generated by the algorithm module and the status information of the exoskeleton human-machine system in the simulation environment to the exoskeleton human-machine system in the real environment to achieve two-way interaction between the real environment and the virtual environment. And synchronization. Different from the traditional simulation environment, based on this simulation synchronization module, the exoskeleton human-machine system model in the real world can synchronize status information with the exoskeleton human-machine system simulation model in the virtual environment to achieve the best understanding of the exoskeleton in the real world. The digitization of the human-machine model facilitates further control and analysis of simulation results; the main function of the simulation synchronization module is to build the corresponding communication protocol based on Ethernet, which is used in the real-world exoskeleton human-machine system and the exoskeleton human in the simulation environment. Perform high-speed data exchange between machine simulation models, transmit real-world data to the simulation environment, and synchronize some data generated in the simulation environment to the real-world exoskeleton human-machine system. In addition, corresponding synchronization mechanisms can also be constructed to ensure that the exoskeleton human-machine system in the real world and the exoskeleton human-machine system in the simulation environment maintain a high degree of synchronization in motion.

Intelligent algorithm module: As shown in Figure 9, this module is the core module of the simulation system. Its main function is to conduct simulation of the exoskeleton human-machine system based on the simulation environment constructed by synchronizing the aforementioned exoskeleton human-machine system to various data in the simulation environment. Perception recognition, cognitive decision-making and control planning algorithm simulation. Among them, the perception and recognition algorithm mainly comprehensively analyzes the collected exoskeleton robot sensing data, wearer data and environmental data, and identifies and analyzes the key information, such as the wearer's physiological state and movement intention, movement status and Stability, terrain environment and size parameters, etc.; the cognitive and decision-making algorithm is mainly based on the results of the perception and recognition algorithm to fully analyze the human-machine motion status data for specific moving targets, and evaluate the exoskeleton human-machine based on the human movement intention and target motion status. Decisions and simulation optimization are made based on the system's subsequent movement trends, travel routes, passing through specific terrain, reaching specific forward speeds, etc. in complex environments, and corresponding prediction results are given; planning and control algorithms are mainly based on cognitive and decision-making algorithms. The prediction results and target motion status are combined with the current exoskeleton human-machine status information to establish a related optimization algorithm to generate subsequent motion planning sequences and generate control instructions in real time. Through the orderly cooperation of the above sub-modules, the intelligent algorithm module can introduce a variety of intelligent algorithms to participate in the perception identification, cognitive decision-making and planning control stages, thereby obtaining different simulation results. In addition, the control instructions generated by this module will also establish communication with the real-world exoskeleton human-machine system through the simulation synchronization module, and control the exoskeleton human-machine system in real time to operate according to the planning results, so that the exoskeleton human-machine model in the simulation environment Achieve a high degree of synchronization with the exoskeleton human-machine system in the real world. The visual interface of the simulation process can be run on a personal laptop/desktop computer/server, and the visual interface can be developed based on software interface development architectures such as QT and MFC. The simulation visualization interface supports mathematical models of customized simulation environments or open source physics simulation engines (ODE, Bullet, Mujoco, etc.). These physics engines are used to build simulation models and their environments, and calculate the kinematics and dynamic motion processes in them. Including the position, speed, acceleration, force of each link and joint in the simulation environment, as well as the collision process, friction process, deformation process of the soft body structure, etc.

Intelligent algorithm modules can be deployed on personal laptops, desktops or servers according to computing needs, or use a distributed multi-point deployment method, with Ethernet-based network protocols used for interconnection in the middle to make full use of the high computing power of the server and personal The mobile flexibility of the computer is coordinated and controlled through the simulation control module in the simulation visualization module.

In the intelligent algorithm module, the perception and recognition algorithms can use conventional pattern recognition and machine learning algorithms to analyze and identify various types of data collected, extract key feature information, improve recognition efficiency, etc.; the cognition and decision-making algorithms can be combined with the wearer's The EEG and EMG signals are used to extract and identify the wearer's movement intention, and are combined with the aforementioned environmental information perception and recognition algorithms to build a collaborative cognitive and decision-making algorithm that integrates humans and machines to improve the intelligence of the entire algorithm. At the same time, it can also realize the human-machine collaborative decision-making strategy of human-in-the-loop; planning and control algorithms can adopt common modeling, control and planning algorithms in the field of robots, or can also be combined with planning control algorithms in emerging robot fields such as imitation learning and reinforcement learning. The human-machine simulation model and environmental model reconstructed in the simulation environment are interactively learned to construct optimization goals to explore optimal planning and control strategies.

Simulation visualization module: As shown in Figure 10, the main function of this module is to synchronously visualize the data generated in the entire simulation system, so that developers can intuitively observe the performance of existing intelligent algorithms implemented on the current exoskeleton human-machine system. Effect. This module can be divided into several sub-modules: human-machine status display and interaction module, data display module, simulation process control module, model synchronization and visualization module. Among them, the human-machine status display and interaction module mainly displays the communication process between the simulation environment and the real exoskeleton human-machine system, including display of connection status, data transmission speed, actual exoskeleton human-machine operating status, etc. And design corresponding interactive controls to facilitate the R&D personnel to operate the on-off and start-stop processes between the simulation environment and the real exoskeleton human-machine system; the data display module is mainly used to display the running process of the real exoskeleton human-machine system and simulation model Various types of data in the system, including posture information, motion information, interaction force and other data, are displayed in the form of data charts and curves to facilitate intuitive analysis of data; the simulation process control module is mainly used to control the operation of the simulation model in the simulation environment. process, including switching and using various intelligent algorithms, starting, stopping and debugging the simulation process, etc.; the model synchronization and visualization module is mainly used for the digital reconstruction of the real exoskeleton human-machine system and its environment in the simulation environment The process is visualized, and the reconstructed exoskeleton human-machine simulation model, terrain environment, etc. are displayed simultaneously with the exoskeleton human-machine system and its environment in the real world. It also displays the current exoskeleton from a first-person perspective. Various information such as the terrain environment, obstacle information, and travel routes in front of the human-machine system facilitates R&D personnel to conduct in-depth debugging of the entire algorithm design and simulation process, improving the effectiveness, feasibility, and stability of the algorithm.

The exoskeleton human-machine simulation model during the simulation visualization process can be used for preliminary structural design through mechanical design software such as Solidworks and AutoCAD to ensure that it is consistent with the exoskeleton real machine model in terms of size, mass, inertia, etc., and can then be exported as URDF through plug-ins. /SDF/XML and other structured robot description files, and import them into the simulation platform to build the corresponding exoskeleton robot model. The size adjustment part of the robot can be adjusted one-to-one according to the current real exoskeleton robot model to ensure that the exoskeleton robot model in the simulation environment is consistent in appearance and size with the exoskeleton robot model in the real world.

Wearers during the simulation visualization process can use mechanical design software such as Solidworks and AutoCAD for early structural design to restore the basic wearer human body model and provide a size adjustment interface to facilitate the later use of wearers with different body sizes in the simulation environment. Refactor.

The terrain environment in the simulation visualization process can be reconstructed in the simulation environment based on the environmental information collected through visual sensing by the exoskeleton human-machine system in the real world and the physics engine used, and be consistent with the real environment. The size is restored one to one, so that the environment faced by the exoskeleton human-machine system can be digitally reproduced in the simulation environment.

During the simulation visualization process, the movement process of the exoskeleton robot and the wearer can be based on the information such as the movement status of the exoskeleton robot and the movement status of the wearer in the real world obtained by the data collection module, and the aforementioned open source physics engine can be used for dynamic modeling and simulation. , and generate corresponding control instructions to control the simulation model to move.

In the simulation visualization module, the movement process of the real exoskeleton robot and the movement process of the exoskeleton human-machine simulation model in the simulation environment, as well as the environmental changes during the process, will be recorded and stored in real time. After the simulation process is completed, the history can be retrieved at any time. Record and playback the simulation process, and simultaneously display various data during the simulation process.

Claims (4)

1. An exoskeleton physical human-machine two-way interactive simulation system, which includes: an exoskeleton human-machine system, a data acquisition module, a simulation synchronization module, an intelligent algorithm module, and a simulation visualization module; The exoskeleton human-machine system includes: an exoskeleton robot, the exoskeleton robot is worn by the wearer, and the skeleton of the exoskeleton robot imitates the limbs below the waist of the human body; The data acquisition module includes: an EEG signal acquisition module, an IMU sensor, a joint angle sensor, an interactive force sensor, a pressure sensor, a depth camera, and an IMU attitude sensor. The EEG signal acquisition module is worn on the human head; the IMU Sensors, joint angle sensors, interactive force sensors, pressure sensors, and depth cameras are all installed on the exoskeleton robot. The IMU attitude sensor is installed on the wearer of the exoskeleton robot. The IMU sensor detects the attitude of the exoskeleton, so The interactive force sensor detects the force exerted by the wearer on the exoskeleton, the pressure sensor detects the pressure on the soles of the exoskeleton, the IMU posture sensor detects the posture of each part of the human body, and the depth camera collects the road conditions in front of the wearer; The simulation synchronization module establishes a data exchange channel between the exoskeleton human-machine system in the real environment and the simulation model in the virtual environment, and establishes communication between the simulation environment and the physical human-machine system through an Ethernet-based network communication protocol. protocol and transfer data between the two; The intelligent algorithm module includes: perception and recognition module, cognition and decision-making module, planning and control module; wherein the perception and recognition module is used to identify the terrain environment, human-machine motion status, etc. from the acquired sensor data. and analysis; the cognition and decision-making module is used to make decisions on subsequent motion paths based on the current environment and motion state of the human-machine system; the planning and control module is used to plan the motion mode of the human-machine system based on the aforementioned cognitive decision-making results And control the human-machine system to complete the movement; The simulation visualization module includes: a data display module, an algorithm simulation visualization module, and a simulation process playback module; the data display module displays all data from the exoskeleton physical human-machine system and intermediate state data generated during the simulation process; algorithm simulation visualization The module displays the simulation results on the future motion state of the exoskeleton human-machine system in the simulation environment; the simulation process playback module plays back all completed historical simulation results and the motion process of the real exoskeleton physical human-machine system. 2. An exoskeleton physical human-computer two-way interactive simulation system according to claim 1, characterized in that the specific calculation steps of the calculation method in the perception and recognition module are: The perception and recognition method consists of a center of mass estimator, a slope estimator and a Hill muscle model; the center of mass estimator obtains the motion state and stability of the human-machine system through changes in its center of mass when it moves; the slope estimator estimates the slope in the external environment, Use this for gait planning on external ramps; First of all, in this exoskeleton robot, the ankle joint is set as a passive joint that can move within a small range in order to better assist walking. Therefore, there are five active joints, namely 2 knee joints, 2 hip joints, and pelvic joints. joints; while the wearer considers the joint angles of 10 active motion joints, namely 2 knee joints, 2 hip joints, 2 shoulder joints, 2 elbow joints, lumbar joints, and pelvic joints; each joint is set There is a local coordinate system, and the world coordinate system is set outside the human-machine system; First, calculate the center of mass position of the human-machine system through the collected data; the formula for determining the center of mass of the human-machine system is as follows: Among them, C _machine is the center of mass position of the human-machine system, is the displacement vector from the human pelvic joint to the origin of the world coordinate system, M _h and M _e represent the masses of the wearer and the exoskeleton robot respectively, M is the total mass of the human-machine system, C _h and C _e represent the wearer and the exoskeleton robot respectively. The center of mass position of the exoskeleton robot, ε is the designed physical coupling model; where the displacement vector /> It is obtained from the data obtained from the IMU attitude sensor on the wearer; the formula for determining the position of the wearer's center of mass is as follows: H＝[H ₁ H ₂ …H ₁₀ ] is the wearer’s transfer matrix, where And so on;/> is the uniform transition matrix between adjacent joints,/> with/> are the rotation matrix and displacement vector from joint i to joint j respectively, where the rotation matrix is obtained by the joint angle measured by the joint angle sensor on the wearer, and the displacement vector is obtained by the attitude sensor on the wearer;/> are the kinematics and static parameters of the human body, which are obtained by collecting the static postures of 30 wearers and using formula (2); The formula for calculating the center of mass position of the exoskeleton robot is as follows: E＝[E ₁ E ₂ …E ₅ ] is its state matrix, where and so on; are its kinematic and static parameters; since the exoskeleton robot only considers five active motion joints, its transfer matrix and kinematic and static parameters are fifth-order matrices; the formula for calculating the physical coupling model ε is as follows: ε＝R _Y (θ _lh )δ ₁ +R _Y (θ _lk )δ ₂ +R _Y (θ _rh )δ ₃ +R _Y (θ _rk )δ ₄ , (4) Among them, θ _lh , θ _lk , θ _rh , and θ _rk are the joint angles of the left hip, left knee, right hip, and right knee respectively. R _Y (θ _i ) represents the rotation matrix of joint i relative to the y axis. Through the joint angle The angle data of the sensor is obtained; δ _j is the weight, j = 1, 2, 3, 4, which is obtained through experiments from the data of the exoskeleton foot pressure sensor; After calculating the center of mass position of the human-machine system through the above steps, the exoskeleton human-machine system is simplified into a two-dimensional linear inverted pendulum model, and the potential energy of the inverted pendulum model during the motion is calculated through formula (5); Among them, x and represent the position and velocity information of the center of mass in the horizontal direction respectively, z _c represents the height of the center of mass, and g represents the gravitational acceleration constant; Orbital energy is used to represent the motion state of the center of mass. During a single swing of the inverted pendulum, the orbital energy remains at a constant value until the next time the supporting legs of the inverted pendulum are switched, and the orbital energy will be recalculated; when the human-machine system moves The current orbital energy E _current is calculated by formula (5); different orbital energies lead to different next steps on the horizontal ground, so there are two situations here. When E _current > 0, it means the next step is completed. After the support legs are switched, the inverted pendulum model can continue to move forward; when E _current <0, it means that the inverted pendulum model does not have enough energy to cross the highest point of potential energy and continue to move forward, and the speed drops to zero before reaching the highest point of potential energy. and starts to move backward; the slope estimator is used to estimate the slope of the slope where the current exoskeleton human-machine system is located in real time, and is built based on the sensing data and geometric size information of the exoskeleton itself; the IMU sensor located in the exoskeleton backpack It can sense the posture information of the entire torso in real time, and the angle sensors located in the hip, knee, and ankle joints can sense the angle information of each joint in real time; It is assumed that the sole of the supporting leg is always in contact with the slope and does not slide relative to it. Therefore, the slope estimator calculates the slope based on the joint angles of the hip, knee, and ankle of the supporting leg and the posture information of the trunk, and the left and right sides The legs are alternately supported; when the exoskeleton human-machine system walks on the slope, it is divided into four situations: (a), (b), (c), and (d) according to the different gait phases of the supporting leg and the swinging leg. Among them: (a) indicates the moment when the swing leg just leaves the ground during the uphill process; (b) indicates the moment when the swing leg is about to touch the ground during the uphill process; (c) indicates the moment when the swing leg just touches the ground during the downhill process. The moment when the leg leaves the ground; (d) indicates the moment when the swinging leg is about to touch the ground during the downhill process; combining these four situations, it is obtained that during walking uphill and downhill, when one leg acts as a supporting leg and the other leg As the posture information of all joint angles and the trunk when the swing leg moves forward in the air, the slope of the current ramp is calculated based on this information and the geometric size information of the exoskeleton human-machine system; assume that θ represents the slope of the ramp, and Defining the θ angle is in a similar coordinate system to the joint angle of the exoskeleton, and also follows the right-hand rule, that is, θ<0 during the uphill process, θ>0 during the downhill process, and θ=0 on the level ground. ; Among them, α ₁ , α ₂ , α ₃ , and α ₄ respectively represent the interior angles corresponding to the vertices A, B, C, and E in the polygon ABCEFA. Point A represents the location of the hip joint, point B represents the location of the knee joint, and point C represents At the location of the ankle joint, point E represents the intersection of the extension line of the calf link CD and the X-axis, and F represents the origin of the coordinate system X-0-Y. There is the following geometric relationship: Based on the plane geometric relationship, the sum of the interior angles of a polygon with N sides is (N-2)·π. For the multi-deformation with 5 sides composed of A-B-C-E-F-A, we get: Based on formula (6), the calculation results of slope θ in four cases are: The slope of the ramp is calculated based on the posture information of the trunk and the joint angles of the supporting legs; The muscle model is shown in formula (9): L ^MT =L ^T +L ^M cosa. (9) Among them, L ^MT is the total length of the tendon and muscle fiber, L ^T is the length of the tendon, L ^M is the total length of the muscle fiber, α is the pennation angle, which is the angle between the muscle fiber bundle and the tendon; the active contraction unit and the parallel elastic unit pass through the feather The shape angle α is connected with the series elastic unit to form a muscle system; the active contraction unit is the active part during the muscle work process, and the parallel and elastic units are the passive contraction part. During the active contraction process, the muscle fibers are attached to the bone surface through stretching. Tendons, thereby driving joint movement; muscle fiber force is the sum of active units and passive units, that is, it is considered that at any time, the total force generated by muscle fibers is F, then: F＝F _CE +F _PE , (10) In the formula, the force generated by the F _CE active contraction unit is the active force, and the force generated by the F _PE parallel contraction unit is the passive force; in the muscle model, F _CE represents the force generated by the chemical reaction inside the muscle cross bridge, which can be considered as Produced by the active contraction of muscle fibers, its size is related to the degree of muscle activation, muscle length, and muscle contraction rate; the specific expression of F _CE is: F _CE =a·F ₀ ·f(l)·f(v), (11) In the formula, a is the degree of muscle activation, which is related to the amplitude of the electromyographic signal; if the activation degree of a certain muscle at a certain moment is 0, it means that there is no active muscle force in the current muscle, then F _CE = 0; the higher the degree of activation. , indicating that the more muscle fibers participate in exercise contraction, when muscle activation a=1, it is considered that all muscle fibers in the target muscle participate in exercise, and the tension generated by the muscle reaches the maximum value at this time; F ₀ is the maximum contraction force of the muscle, and its size is affected by The influence of cross-sectional area on muscle physiological morphology: F ₀ =K·PCSA, (12) In the formula, PCSA is the cross-sectional area of the muscle, and K is the maximum stress that the muscle can provide. The length of muscle fibers and the rate of muscle contraction are related to the angle of the moving joint; The force generated by the F _PE parallel contraction unit is the passive force, which is generated by the elastic deformation of muscle fibers. When the length of the muscle fiber exceeds a certain range, the muscle will generate tension to force the muscle fiber to return to its original state. Therefore, the size of F _PE is mainly affected by the maximum contraction force (F ₀ ) dynamic length relationship: F _PE =F ₀ ·f(l). (13) After calculating the joint muscle strength, by comparing it with the muscle strength data of healthy volunteers and the wearer's muscle strength data, the physiological state of the wearer at this time can be obtained; in active training, it can also be used as output Extract and identify the wearer's active movement intention. 3. An exoskeleton physical human-computer two-way interactive simulation system as claimed in claim 1, characterized in that the specific calculation steps of the calculation method in the cognitive and decision-making module are: When the environment is flat, it is assumed that the process of exchanging supporting legs of the inverted pendulum model is completed instantaneously without losing energy. The next step is calculated based on the center of mass position, center of mass velocity and given orbital energy of the current inverted pendulum model; Assume The orbit energy of the inverted pendulum model after switching the supporting legs is E _desired , then: where x _t and Respectively represent the position and acceleration of the center of mass in the horizontal direction at the next time the support leg is switched, x _cp represents the next foothold on the horizontal ground, and z _c represents the height of the center of mass; Therefore, given the desired orbital energy E _desired , the orbital energy adaptive capture point x _cp is calculated as: Among them, the sign() function is a sign function, when The return value is 1, otherwise the return value is -1; the switching time of the supporting leg should be a tested experience value, and the selected switching time must ensure/> The above formula is meaningful, otherwise the orbital energy cannot reach the _desired Edesired in the next step, and multiple steps are needed; given a desired orbital energy _Edesired , each step is calculated through the current motion state of the center of mass. A foothold for one step; When there is a slope in the environment, the equation of the straight line is: z＝z ₀ +kx, (16) Among them, the slope k=tanθ, θ is the slope of the ramp, z ₀ represents the height of the center of mass at the initial moment; it is worth noting that during the uphill movement, θ<0, so k<0; during the downhill movement, θ>0, therefore k>0; The way to calculate the motion state of the center of mass is: in is a constant, t represents time; in addition, the movement in the vertical direction is obtained by the straight-line formula of the center of mass moving along the ramp. The calculation method is as follows: z _t =z ₀ +k·Δx＝z ₀ +k·(x _t -x ₀ ) (18) Among them, z ₀ represents the height of the center of mass at the initial moment, and Δx represents the horizontal movement distance of the center of mass from the initial moment to the moment when the supporting leg switches; then, assuming that the inverted pendulum model is in a certain swing process, the position of the swinging legs at the initial moment is _P0 : Then when the supporting leg is switched at time t, the foothold of the swing leg is calculated using formula (20): The starting position P ₀ of the swinging foot and the next landing position P _cp have been obtained. Based on these two positions, a spatial trajectory from the starting position of the swinging foot to the landing point is planned as the gait trajectory of the exoskeleton robot. 4. An exoskeleton physical human-computer two-way interactive simulation system as claimed in claim 1, characterized in that the specific calculation steps of the planning and control method are: The basic model for modeling the gait trajectory of a lower-limb walking exoskeleton robot is: where τ is a constant, used to scale the amount of time to adjust the speed of the trajectory, α _y , β _y are both constants, y=[xz] ^T represents the spatial position of the swinging foot, represents its speed and acceleration, y _cp represents the final expected foothold; f is a nonlinear term, as shown below: N represents the number of basis functions, v comes from a first-order system, and satisfies where a _v is a constant; the basis function Ψ _i is the radial basis function: Among them, σ _i and c _i respectively represent the width and center value of the basis function Ψ _i ; assuming that the teaching trajectory is expressed as Then we get the f _target that needs to be learned: To determine the weight value w _i corresponding to each basis function Ψ _i , construct a loss function: Among them, t represents the time step, and T represents the total number of time steps of the entire trajectory; the minimization process of solving the above loss function is to obtain the parameters of the model. After the gait trajectory model learning is completed, different initial positions and end positions of the trajectory are given. to generate different gait trajectories.