39 results about "Gait planning" patented technology

The invention discloses a humanoid robot motion control method based on gait planning and reinforcement learning, and belongs to the technical field of humanoid robot motion control. According to the method, the problems of low convergence speed in a training process and poor stability and reliability of a trained model of an existing reinforcement learning method are solved. According to the method, a reinforcement learning model runs in parallel under multiple environment instances, in the running process, a target foothold is generated in real time based on feedback state information, the generated reference foothold and a mass center track serve as an action reference or reward target of a controller, and a continuous and dynamically-adjusted foot end track is constructed in combination with gait phase information. And after the reinforcement learning model is trained by utilizing all the collected trajectory data, motion control can be performed on the humanoid robot through the trained reinforcement learning model, so that the robot realizes stable walking similar to human beings under the condition of not depending on a predefined template. The method can be applied to motion control of the humanoid robot.

The invention belongs to the related technical field of biped robot gait planning, and discloses a method, application and equipment for generating a biped robot anthropomorphic foot end trajectory based on a full-connection network, and the method comprises the steps: (1) grouping the collected foot end trajectory data according to a starting step, an intermediate circulation step and a termination step; (2) mapping each group of foot end track data into foot end track data which can be used by a robot based on a kinematics mapping relation between a human body and the robot, and carrying out conformal interpolation; (3) training a full-connection neural network model by adopting the interpolated foot end trajectory data, further generating a foot end trajectory of each intermediate step by adopting the full-connection neural network model, and mapping the directly generated foot end trajectories of the starting step and the ending step to jointly form an anthropomorphic foot end trajectory; and the main loss of back propagation of the full-connection neural network model is punished by introducing the trajectory smoothness and the prediction precision of the start end and the tail end. According to the method, the anthropomorphism of the foot end track is improved.

The invention discloses a small robot hierarchical reinforcement learning gait adaptive generation method and system, and the method comprises the steps: collecting environment perception data and robot body state data through a perception module, and extracting environment features and robot state information; generating a gait plan instruction through a high-level planning module based on the environment characteristics and the robot state information; based on the gait plan instruction and the robot state information, a joint control instruction is generated through a low-layer control module in combination with a multi-body dynamics model; the joint control instruction is executed, and robot joints are controlled to move so as to achieve self-adaptive gait generation. By utilizing the embodiment of the invention, the problem of self-adaptive generation of gaits in a complex environment can be solved, and the motion intelligence, the control accuracy and the quick adaptability to unknown terrains of the robot are improved.

The invention relates to a biped robot gait planning method. The invention relates to the technical field of robot gait planning, and the method comprises the steps: taking a kinetic model of a robot as input, so as to achieve the neural network reasoning of simulation; describing the robot by using a URDF file, and enabling the robot to be equivalent to a kinematic chain formed by connecting rigid bodies in series; through accurate description, a robot consistent with that in reality is established in a simulation environment, the robot is enabled to carry out gait learning in a simulation physical environment, and supervised random divergence trial and error are carried out in a mode of setting initial conditions, termination conditions and reward and punishment conditions so as to realize gait learning. According to the invention, through repeated modification of parameters and structural adjustment of adaptability, a stable, low-energy-consumption and high-robustness control strategy is finally obtained through training.

The application discloses a kind of small celestial body weak gravity condition under detection robot movement control method and device, the method of the present application includes first to detection robot is carried out steady gait planning, at the end time of the foot end trajectory obtained, add the active attachment bias Δr for ensuring that detection robot is attached to the clamping of small celestial body star table;Then the foot end trajectory is as leg desired position, inverse kinematics is solved to obtain the desired joint angle of each joint, obtains the current joint angle of each joint of detection robot, and calculates the deviation between the current joint angle and the desired joint angle of each joint, and the deviation is input into the movement controller to obtain the joint output control torque, the corresponding joint is controlled by joint output control torque, until the deviation is less than set value.The present application can realize the stable attachment and movement of detection robot on the surface of small celestial body under the condition of weak gravity of small celestial body, create prerequisite for the smooth development of detection task.

The application discloses a lower limb rehabilitation training exoskeleton control method, an electronic device and a storage medium. The method comprises the following steps: collecting standard gait information in the process of human walking; generating a hip joint angle change curve and a knee joint angle change curve according to the standard gait information; performing smoothing processing on the hip / knee joint angle change curve; generating a variable step frequency step length hip joint curve according to the hip joint angle change curve and a hip joint amplitude parameter and a step frequency parameter; generating a variable step frequency step length knee joint curve according to the knee joint angle change curve and a knee joint amplitude parameter and a step frequency parameter; and driving a hip / knee joint motor according to the variable step frequency step length hip / knee joint curve, and driving a human lower limb through an exoskeleton. The method is based on normal motion data of the human body for gait planning, is closer to the human motion mode, has strong adaptability, requires a small amount of data, has strong practicability, and solves the problem of speed mutation caused by non-smooth adjacent period curves.

This invention discloses a hierarchical reinforcement learning method and system for adaptive gait generation in small robots. The method includes: collecting environmental perception data and robot body state data through a perception module, and extracting environmental features and robot state information; generating gait planning instructions through a high-level planning module based on the environmental features and robot state information; generating joint control instructions through a low-level control module combined with a multibody dynamics model based on the gait planning instructions and robot state information; and executing the joint control instructions to control the robot joint movement to achieve adaptive gait generation. Using this invention, the problem of adaptive gait generation in complex environments can be solved, improving the robot's motion intelligence, control accuracy, and rapid adaptation to unknown terrain.

The application relates to the technical field of robot simulation, in particular to a robot ground verification platform and a robot vibration simulation method. The robot ground verification platform is used for verifying the vibration response of a space robot walking on a flexible structure and performing gait planning, and comprises a platform main body, a flexible structure beam, a fixed base, an electromagnet, an acceleration sensor, a simulated floating wheel and a robot. The flexible structure beam is arranged on the platform main body through the fixed base. The acceleration sensor and the electromagnet are arranged on the flexible structure beam. The robot moves on the platform main body through the simulated floating wheel. The robot vibration simulation method is that the robot moves on the flexible structure beam through the on-off state of the electromagnet, and the vibration of the flexible structure beam is detected through the acceleration sensor. The application can simulate the action of the robot in a space state, can monitor the vibration of the flexible structure beam in real time, so that the vibration parameters of the robot can be obtained, and the application of the experiment is facilitated.

The invention provides a wheel step type rotary drilling rig intelligent gait planning and cooperative control method and system, and the method comprises the steps: collecting the torque value of a drill rod, the rotation direction of the drill rod, the contact force deflection angle of each supporting leg steel claw, the leg vibration frequency, and the feeding speed and rotation speed of drilling operation; fourier transform is carried out on the torque value of the drill rod to obtain the torque fluctuation frequency of the drill rod; evaluating a phase angle between the rotation direction of the drill rod and the circumferential layout of the supporting legs, screening out leg positions with the phase angle exceeding a preset safety angle threshold value and vibration exceeding a limit, and adjusting the unfolding radius; identifying load concentration orientation distribution of each leg by taking the expansion radius adjustment amount of the supporting leg as an independent variable and the steel claw contact force deflection angle change rate as a dependent variable; gait scheduling arrangement, feeding speed and rotation speed are fused and optimized, a switching time window is divided according to the leg lifting and falling action period of each supporting leg, and feeding propulsion and rotation start-stop takt is matched in an alternating gap of adjacent supporting legs to form a cooperative control signal.

The application relates to an exoskeleton robot gait planning method, device, system and storage medium. The method comprises the following steps: acquiring angle data of a human body wearing an exoskeleton robot in a current walking cycle; the angle data is output by a motor encoder at a joint position of the exoskeleton robot; performing time domain to frequency domain conversion on the angle data to obtain frequency domain data; determining a gait planning function based on a gait fitting function, gait fitting parameters and the frequency domain data, and outputting the gait planning function; and the gait planning function is used for gait planning of the exoskeleton robot. The method can effectively improve the matching degree of the motion trajectory of the exoskeleton robot and the motion trajectory of the human body.

The invention discloses a safety constraint optimization gait planning method and device for a lower limb exoskeleton rehabilitation robot, and the method comprises the steps: designing an ACPG (adaptive central pattern generator) network topology for sagittal plane activity space limitation, and achieving the generation of a personalized gait track with a biological rhythm under human-computer interaction; on the basis of interaction torque and joint speed fed back by sensing, energy of human body active movement injected into the exoskeleton is calculated; constructing an intermediate variable according to the energy weighting, designing a gait adjustment model in combination with an admittance model, and outputting a self-adaptive smooth gait frequency and an amplitude modulation factor; based on the gait adjustment model, by using a control barrier function optimization theory, gait frequency and amplitude change are constrained within a safety range, so that the ACPG outputs rhythm gaits and meets safety requirements at the same time. According to the method, the energy change under the human-computer interaction effect is considered, and the personalized gait track with safety protection is generated.

The application relates to the field of robot control, and particularly discloses a PID control and gait planning method of a wheel-foot composite structure and electronic equipment, which comprises the following steps: a PID control model with a coupling compensation term is established; each joint angle error is calculated; adjacent joint motion interference is compensated according to the PID control model with the coupling compensation term; in the wheel mode, an outer ring controls speed with the second threshold value as the target, and an inner ring corrects the posture according to IMU feedback; when the inclination exceeds the third threshold value, the posture control weight is enhanced; in the foot mode, PID parameters are adjusted in advance according to laser radar prediction, and the adaptability to complex terrains is improved; terrain, gait, control precision and energy consumption data are collected in real time, PID parameters are automatically adjusted based on a PPO algorithm, and the control precision of the robot in unknown terrains is improved; and through a dynamic coupling compensation mechanism, the motion interference between joints is accurately resolved, each joint can keep high coordination in motion, and accurate motion postures can be stably maintained regardless of complex terrains.

The invention relates to a self-adaptive gait planning method and system based on whole body motion collaborative optimization, and belongs to the field of robot control. The method comprises the following steps: obtaining virtual terrain impedance through multi-modal sensing fusion; dynamic CPG modulation is carried out based on the virtual terrain impedance, an adaptive gait rhythm is obtained, a basic gait mode is generated, and the adaptive gait rhythm comprises a leg phase, gait amplitude and gait frequency; the basic gait mode is corrected based on whole body movement collaborative optimization, and an ideal foothold and an expected mass center acceleration vector are obtained. According to the method, virtual terrain impedance perception, dynamic CPG modulation and whole-body motion collaborative optimization are fused, unified processing of terrain static attributes and dynamic disturbance is achieved, and the adaptive capacity and robustness of the robot in an unknown environment are remarkably improved.

The invention discloses a nerve-mechanics coupling software multi-mode motion control method based on a pre-training language model. Comprising a mass point-spring-based soft robot body structure module, a rhythm oscillation signal generation and parameter control module modulated based on a pre-training language model, a muscle driving and mechanical calculation module based on nerve-mechanical mapping, and a simulation and debugging module based on physical simulation. The method is oriented to gait planning and intelligent control of the bionic soft robot in an unstructured environment. In order to solve the problem that gait switching depends on artificial parameters and is difficult to quickly adjust along with instructions, a pre-training language model is introduced, and high-level semantic instructions are mapped into neural controller characteristic parameters. The robot can smoothly switch wriggling, steering and rolling behaviors under the working conditions of a hard plane, viscous fluid and the like through a natural language without remodeling, and a control and analysis method for effectively improving the autonomous environment adaptability of a software system is provided.

The application discloses a disturbance estimation and humanoid robot whole-body disturbance rejection control method based on DCM residual, and relates to the technical field of humanoid robot motion control. A disturbance observer based on DCM residual is designed, external force disturbance acting on the robot centroid can be estimated in real time without the need of a foot force sensor, and the disturbance observer has stronger robustness to model errors and sensor noise; secondly, a hierarchical online motion planner is proposed; a foothold planner optimizes the foothold position and gait cycle in real time through quadratic programming; a gait generator adopts a model predictive controller based on three-dimensional DCM, generates a robust centroid trajectory and gait according to the planned foothold; further, a whole-body disturbance rejection controller is developed, the estimated disturbance is embedded into the whole-body dynamics constraint, active compensation of the disturbance and direct control of the joint torque are realized. The application is used to solve the problems of insufficient robustness, low disturbance estimation accuracy and coupling of gait planning and whole-body control existing in the prior art.

The invention provides a steel gate detection six-foot three-section type wall-climbing robot and a truss crossing method thereof.The steel gate detection six-foot three-section type wall-climbing robot comprises a middle body module and end body modules at the two ends, first supporting legs are arranged on the two sides of the front end body module, second supporting legs are arranged on the two sides of the middle body module, and third supporting legs are arranged on the two sides of the tail end body module; two overturning modules are arranged between the end body modules and the middle body modules, detection and positioning modules are arranged on the end body modules, and the end body modules are rotationally connected with the middle body modules. According to the robot, the six-foot three-degree-of-freedom series configuration design is adopted, each leg module can independently achieve multi-dimensional movement, rotation around the Z axis, pitching around the X axis and the like, the robot has the bionic gait planning capacity, the foot falling point and the supporting posture can be flexibly adjusted, and therefore the robot stably crosses discontinuous and high-fall obstacles which a traditional moving mechanism difficultly passes through. The segments are connected through the high-rigidity overturning modules, and the problem that an integral robot is prone to instability and backward inclination when climbing over tall and large obstacles is solved.

This invention relates to a phased adaptive control method for an exoskeleton used in rehabilitation training, belonging to the field of human-machine collaborative motion assistance technology. This method identifies gait phases based on hip joint angle and angular velocity, dividing the gait into four phases: early swing phase, late swing phase, early stance phase, and late stance phase. It then uses a polynomial curve with control points to complete basic gait planning. Finally, it adaptively adjusts the gait trajectory control points based on the swing dynamics and stance phase. This invention enables phased adaptive adjustment of gait trajectory and auxiliary torque, improving training flexibility and patient participation, effectively avoiding sensor signal interference and insufficient accuracy, and contributing to improved rehabilitation training outcomes.

The invention discloses a steel box girder crack fluorescence labeling robot device and a method thereof, relates to the field of bridge detection, and aims to solve the problems that hidden cracks in a steel box girder are difficult to inspect and high in cost. The device comprises a fluorescence labeling system integration module, a motion module and a main control module, the motion module is of a four-foot magnetic adsorption structure, wall face vertical transition gait planning and cooperative control of driving magnetic force, a motor and gaits can be achieved, and the robot has the obstacle crossing function and is suitable for narrow and complex terrains. The fluorescence labeling module adopts a high-precision spray head and ultraviolet light to excite paint, the laser sensor scans feedback data in real time, and the master control module identifies cracks and plans a spraying path. The method comprises the steps of parameter presetting, scanning recognition, gait and magnetic force adjustment, mark spraying and circulating operation. Marking is clear and accurate, the labor cost is greatly reduced, the bridge maintenance safety and intelligent level are improved, and the method is suitable for efficient marking of steel box girder cracks.

The invention provides a slope terrain gait planning method and device for an underwater quadruped robot, and relates to the technical field of robot control, and the method comprises the steps: obtaining the trim angle of a robot body and a slope tilt angle, and determining the maximum trim angle of the robot on a current slope; whether the maximum trim angle is larger than or equal to the slope inclination angle or not is judged, if yes, a preset gait is adopted for climbing, and if not, the gait of the robot is dynamically adjusted; for the supporting legs, minimization of the tangential force of the front supporting legs and approaching of the propelling force of the rear supporting legs to a target value serve as optimization targets, under the condition that force and moment balance, friction cone constraint and pressure constraint are met, target foot end force of all the supporting legs is dynamically distributed, and control instructions of corresponding joints of all the supporting legs are executed; for the swinging legs, the gait cycle is prolonged according to the difference value between the slope inclination angle and the maximum trim angle, the swinging height of the front swinging leg and the falling point position of the rear swinging leg are adjusted in a self-adaptive mode, and control instructions of the joints corresponding to the swinging legs are executed.

The invention discloses a four-footed robot gait planning method based on Stateflow control, which belongs to the field of kinematics of four-footed crawling robots, and comprises the following steps: modeling a four-footed robot, and setting geometric parameters of the four-footed robot; gait planning of the quadruped robot is designed, and gait control parameters are set; motion equation analysis is carried out based on the gait control parameters, and a foot end gait curve of the quadruped robot is constructed; establishing a relationship between stretching of two hydraulic cylinders at foot ends and rotation angle changes of thighs and shanks; establishing a relation between foot end coordinates of the quadruped robot and rotation angles of thighs and shanks; constructing a Stateflow system as a control program for movement of the quadruped robot; and simulating the movement of the quadruped robot based on gait planning and a foot end gait curve. According to the four-foot robot motion control method, the motion process of the robot is planned by means of dynamic analysis, the impact on the feet of the robot is reduced, and the flexibility and adaptability of four-foot robot motion control are remarkably improved.

This invention discloses a hierarchical collaborative reinforcement learning motion control architecture for bipedal robots. First, a sensing unit collects real-time robot state and environmental data, providing rich information to the upper-level main control unit to assist in making rational decisions and improve adaptability to complex terrains. Second, the upper-level main control unit uses reinforcement learning models for inference, enabling gait planning, motion pattern decision-making, and terrain recognition. It can automatically adjust according to terrain and task, improving stability and flexibility. It can also output control parameters for adaptive control, dynamically adjusting based on multiple factors to meet complex motion requirements. Furthermore, the hierarchical collaborative architecture features clear division of labor and collaborative cooperation, efficiently handling control problems and improving system performance and response speed. Finally, the middle-level computation unit dynamically allocates computation tasks according to the computing power of heterogeneous chips, fully utilizing chip potential, improving computational efficiency, reducing computational latency, enhancing the robot's real-time response capabilities, and better adapting to complex environments and tasks.

The present application relates to the technical field of robot control, in particular to a kind of water reservoir bottom mud dredging robot gait planning method and system, comprising: based on the collection multimodal sensing information, constructs the environment semantic map of dredging robot;According to the environment semantic map, drive foot end actuator and water reservoir bottom mud environment interact, and according to the real-time force and motion data existing in the interaction process, obtain the local environment dynamics parameters of water reservoir bottom mud environment;According to local environment dynamics parameters and preset gait sequence instruction, predict the predicted joint control instruction on future period, and drive foot end actuator according to predicted joint control instruction;The execution parameter of foot end actuator is monitored, and when execution parameter exceeds preset safety execution parameter, generate fault-tolerant joint control instruction to drive foot end actuator, effectively solve the technical problems that existing gait planning method cannot guarantee the stable operation and efficient operation of dredging robot.

The present application belongs to the technical field of mobile robots, and discloses a four-wheel-foot robot foot type motion gait planning method and a four-wheel-foot robot, wherein the method comprises the following steps: S1, in the swing period of a swing leg, the real-time center of gravity position of the robot body is calculated and obtained; S2, under the condition that the motion trajectory of the robot body is unchanged, the support position of the support leg is adjusted through the rolling of the wheels according to the real-time center of gravity position, so that the projection of the real-time center of gravity position on the ground is located in the support triangle formed by the support position of the support leg. The present application considers that when one leg of the four-wheel-foot robot swings, the wheels at the bottom of the remaining three support legs can provide support force to the robot body while moving, and based on this, it is proposed that when the four-wheel-foot robot travels in the foot type motion mode, the wheel foot in the support phase adjusts the support position through the rolling of the wheels, so as to expand the motion stability domain, which is beneficial to improve the terrain adaptability and obstacle crossing performance of the robot.

The utility model provides a tandem wheel-leg robot, and belongs to the technical field of machinery. Comprising a frame body, and hip joints are fixedly connected to the two sides of the frame body; and the leg connecting rods are used for improving the energy transmission efficiency of the robot, and the leg connecting rods are connected with the hip joints. The leg connecting rods are arranged, the leg connecting rods are of a four-connecting-rod structure, the length and the angle of the connecting rods can be adjusted through mutual cooperation of the first connecting rods, the second connecting rods and the third connecting rods, and therefore complex gait planning can be achieved, and different movement requirements can be met; the robot can flexibly adjust the posture and the gait in a complex environment, has the advantages of being simple in structure, high in stability, diversified in motion form, low in energy consumption and the like, and can remarkably improve the performance and the adaptability of the robot.

The application discloses a control method and system of an omnidirectional terrain adaptive wheeled robot chassis, and comprises the following steps: receiving a task instruction; performing path planning to generate a plurality of preset action template combinations of the wheeled robot moving from an initial position to a target position and posture change parameters in the moving process; performing gait planning to obtain gait planning instructions; performing inverse kinematics calculation on the posture change parameters to obtain posture control instructions; performing superposition calculation on the gait planning instructions and the posture control instructions to obtain motor target angles in each execution unit of the wheeled robot for chassis moving control; obtaining a position relationship of a current vehicle body of the wheeled robot relative to a map coordinate system to perform path planning adjustment and gait planning adjustment; and obtaining angles of the current vehicle body of the wheeled robot relative to three axes of an earth coordinate system to adjust the posture control instructions. The application guarantees chassis tracking performance and motion performance, and can adapt to more extensive terrain environments and load requirements.

The invention relates to a gait planning technology of a humanoid robot, in particular to a humanoid robot steering gait planning method based on nonlinear model predictive control. The method provided by the invention is different from the prior art, not only realizes planning of the gait sequence, but also can plan the robot direction angle track meeting the requirements at the same time. According to the gait planning method, the orientation angle can be autonomously adjusted according to the expected speed of an upper-layer instruction, so that the robot has better trajectory tracking performance during steering. Meanwhile, different from the previous method for conservatively approximating the double-support-period ZMP constraint to the single-support-period constraint, the method disclosed by the invention explicitly processes the double-support-period ZMP support polygon constraint in the model prediction control framework, expands the support domain of the optimization problem, enhances the solving flexibility and success rate of the algorithm in a complex dynamic scene, and improves the optimization performance.

The application provides a low-resistance gait planning method, device and equipment for an underwater robot and a storage medium, and relates to the technical field of robot control. The method comprises the following steps: dividing the motion gait of a legged robot into a support phase and a swing phase. In the support phase, the legged robot is propelled forward by the reaction force of the contact between the foot end and the ground. In the swing phase, the legged robot moves forward in the posture of retracting the limbs to reduce the frontal area by imitating the frog swimming. The spatial position of the foot end of the legged robot in the trunk coordinate system is calculated according to the rotation angle of each joint. The real-time resistance of the legged robot is calculated based on the real-time frontal area of each leg segment of the legged robot and the real-time speed relative to the water flow. A prediction model is constructed, a target function and a constraint condition are set, a sequential quadratic programming algorithm is used for rolling optimization and solving, and the optimal torque sequence of each joint is obtained. The optimal torque sequence is sent to the joint motor for execution, so as to control the low-resistance gait motion of the underwater robot.

The invention discloses an operation method of a front elbow and rear knee two-stage motor-driven quadruped robot, the quadruped robot comprises a trunk, a left forelimb, a right forelimb and other structures, the left forelimb and the right forelimb are arranged on the front portion of the trunk and are both of an elbow type structure, and the operation method of the quadruped robot comprises the steps that S1, according to the preset operation requirement of the quadruped robot, the left forelimb and the right forelimb are connected; determining the time proportions of the supporting phases and the swinging phases of the left foreleg, the right foreleg, the left posterior limb and the right posterior limb in one gait cycle. According to different preset operation requirements of low-speed stability, medium-speed high efficiency and the like, accurate matching is realized through differentiated gait planning, a walking gait is adopted for a low-speed scene, and stable movement under a complex terrain is guaranteed; diagonal gaits are adopted in a medium-speed scene, and the exercise efficiency is improved. And meanwhile, matched driving force and adjustment precision are provided for different gaits through cooperative control of the two-stage motor, so that the quadruped robot can flexibly cope with multiple scenes from daily inspection to medium-speed operation.

The invention relates to a crawling gait planning method for a crawling mechanical arm, and belongs to the technical field of mechanical arm motion control and motion planning. Comprising the following steps: configuring a mechanical arm; a starting point, a target point and a movement track of the mechanical arm are given, and grids are divided; calculating an included angle and a distance according to the relative displacement of the first middle point and the starting point; the motion angle of the pitching joint is solved according to inverse kinematics, the mechanical arm moves to the position above a preset connector of a first middle point, then the mechanical arm vertically descends to make contact with the preset connector, the second tail end of the designated mechanical arm grabs the preset connector, and the first tail end of the designated mechanical arm is loosened; the movement angle of the pitching joint is solved again, so that the movement distance of the mechanical arm is achieved, and a working space is reserved for subsequent movement; according to the relative displacement between the second middle point and the first middle point, the vector, the included angle and the distance of the second middle point and the first middle point under the tail end two-coordinate system and the motion mode judgment angle of the mechanical arm are worked out; and judging whether the angle is larger than the reachable angle range of the seventh joint of the mechanical arm.

The application discloses a kind of based on reinforcement learning's six-foot robot fault-tolerant gait planning method, comprising the following steps: build the simulation model of six-foot robot;Establish the CPG gait generator based on improved Hopf oscillator;Fusion simulation model, strategy network and the CPG gait generator based on improved Hopf oscillator, construct reinforcement learning motion control framework, for the established six-foot robot simulation model, fusion reinforcement learning motion control framework;Simulate six-foot robot partial leg failure, freeze six-foot robot leg failure in simulation environment, train reinforcement learning motion control framework;The control network after training is integrated in the gait control framework of six-foot robot, for generating the fault-tolerant gait of six-foot robot and verifying, if it can be completed autonomous motion under the condition of freezing failure leg then it indicates that strategy network is effective, to extract strategy network for controlling the motion of real six-foot robot.