A control method and system for connection transition period of a biped robot based on MPC-CPG fusion

Abstract

The application provides a control method and system for a transition period of a biped robot based on MPC-CPG fusion, which comprises the following steps: calculating a corresponding overall center of gravity position in a transition stage; initializing four Hopf oscillator CPG networks, and starting a corresponding time-varying weight function to generate CPG initial states of four foot ends and continuous foot end reference trajectories; in each control cycle, corresponding ZMP positions, overall center positions and CPG states of the two biped robots are updated to build a corresponding MPC prediction model; a target function is solved by using the MPC prediction model to output first-step optimal CPG parameters of the two biped robots and generate corresponding joint instructions; after it is judged that the two biped robots meet a transition end condition, the transition stage is exited, and a steady-state quadruped CPG planning is switched synchronously. The application can greatly improve walking stability.

Description

Technical Field

[0001] This invention relates to the field of robot control technology, and in particular to a control method and system for the connection transition period of a bipedal robot based on MPC-CPG fusion. Background Technology

[0002] Legged robots, with their superior terrain adaptability and mobility, have become core intelligent equipment for complex scenarios such as industrial inspection, emergency rescue, and wilderness exploration. Among them, bipedal robots have the advantages of navigating narrow spaces and nimbly overcoming obstacles, while quadrupedal robots have higher load-bearing capacity and walking stability. Reconfigurable modular legged robots, through the dynamic combination of multiple bipedal robots, can balance the scene adaptability of bipedal robots with the movement stability of quadrupedal robots, and have become an important development direction in the field of legged robots.

[0003] However, in modular combinations where two independent bipedal robots are rigidly connected to form a quadruped structure, a dynamic balance bottleneck urgently needs to be addressed during the gait transition period (approximately 1.5–3 seconds) after connection. At the instant of connection, the robot's overall center of gravity shifts abruptly, and the quadrupedal support polygon expands instantaneously. The original bipedal gait phase, walking speed, and joint torques of the two robots become severely mismatched with the target quadrupedal gait, directly causing the zero-moment point (ZMP) to rapidly exceed the stability margin, joint torques to abruptly change, and the overall robot posture to fluctuate violently. This problem not only poses a very high risk of tipping over and falling, easily damaging joints and sensor hardware, but also prolongs the transition time and causes roll / pitch angle fluctuations exceeding 10°, preventing the subsequent quadrupedal steady-state gait from starting normally. The stability advantage of modular combination is completely lost, severely hindering the engineering implementation and practical application of reconfigurable legged robots.

[0004] Current technologies for gait control of legged robots mostly employ model predictive control (MPC) for trajectory optimization and central pattern generators (CPGs) based on Hopf oscillators for gait rhythm generation. These solutions only address steady-state walking control or routine gait switching scenarios for a single quadruped robot, without considering the sudden changes in the overall dynamic characteristics of two independent bipedal robots after connection or the specific scenario of collaborative mapping of the dual-robot operating states. They cannot solve the problem of dynamic balance control during the gait switching transition period and are insufficient to meet the hot switching and complex terrain adaptation requirements of reconfigurable modular legged robots. Summary of the Invention

[0005] Based on this, the purpose of this invention is to provide a control method and system for the transition period of bipedal robots based on MPC-CPG fusion, so as to solve the problem that the existing technology cannot solve the dynamic balance control problem during the transition period of robot gait switching, which leads to the difficulty in adapting to complex terrain requirements.

[0006] The first aspect of the present invention proposes: A control method for the connection transition period of a bipedal robot based on MPC-CPG fusion, wherein the method includes: Once the connection between the two bipedal robots is detected, in response to the corresponding connection signal, the IMU data, joint status data, and six-dimensional sensor data of the four feet of the two bipedal robots are collected during the transition phase to calculate the corresponding overall center of gravity position. Initialize four Hopf oscillator CPG networks and start the corresponding time-varying weight function to generate the initial state of the CPG at the four feet and the continuous foot reference trajectory. In each control cycle, update the ZMP position, overall center position and CPG state of the two bipedal robots to construct the corresponding MPC prediction model. The objective function is solved by the MPC prediction model to output the first optimal CPG parameters of the two bipedal robots, and the first optimal CPG parameters are updated to the Hopf oscillator in real time to generate the corresponding joint commands. The joint commands are sent to the two bipedal robots. After determining that the two bipedal robots meet the transition end conditions, the transition phase is exited and the system is switched to steady-state quadruped CPG planning to complete the corresponding control.

[0007] Furthermore, the expression for the algorithm that calculates the corresponding overall center of gravity position is as follows:

[0008] in, This represents the coordinates of the overall center of gravity of the quadrupedal system formed after connection. The masses of the two individual bipedal robots are respectively. These are the coordinates of their respective centroid positions at the current moment.

[0009] Furthermore, the step of initializing the four Hopf oscillator CPG network includes: After entering the transition phase, four mutually coupled improved Hopf oscillator networks are used as rhythmic trajectory generators for the four feet, corresponding to the left front leg, front right leg, rear left leg, and rear right leg of the two bipedal robots, respectively. The leg mapping rules are as follows: the left leg of the first bipedal robot is mapped to the front left leg of a quadruped, and the right leg is mapped to the front right leg of a quadruped; the left leg of the second bipedal robot is mapped to the rear left leg of a quadruped, and the right leg is mapped to the rear right leg of a quadruped.

[0010] Furthermore, the step of activating the corresponding time-varying weighting function to generate the initial CPG states of the four foot ends and the continuous foot reference trajectory includes: In each control cycle of the transition phase, a time-varying weight function is used as a fusion mechanism to smoothly map the original gait state of the two bipedal robots to the initial CPG state of the four feet in real time. A continuous foot reference trajectory is generated based on the mapped initial state of the CPG.

[0011] Furthermore, the expression for the algorithm that smoothly maps the original gait states of the two bipedal robots to the initial CPG states of the four feet in real time is as follows:

[0012] in, quad,i (0) represents the initial phase used to initialize the i-th leg of the quadrupedal CPG after fusion mapping, and w(t) represents the time-varying weight function. biped,orig This indicates the original bipedal gait phase state of the two robots. quad,ref The reference phase representing the target's quadrupedal gait; The expression for the time-varying weighting function is:

[0013] in, Represents the time-varying weighting function. This represents the current time variable after entering the transition phase. Indicates the attenuation coefficient; The expression for the algorithm that generates a continuous foot reference trajectory based on the mapped initial state of the CPG is as follows:

[0014] in, Indicates the reference trajectory of the foot. Indicates the position of the overall center of gravity. Indicates the output signal of the oscillator. Indicates amplitude. Indicates the offset. This indicates the height offset.

[0015] Furthermore, the expression for the MPC prediction model is:

[0016] in, These represent the system state transition matrix and the control input matrix, respectively. Represents the control input vector. The external disturbance term received by the system is represented by x(k), the state vector is represented by x(k+1), and the MPC prediction model is represented by x(k+1).

[0017] Furthermore, the expression for the objective function is:

[0018] in, This represents the objective function for MPC rolling optimization. Indicates the number of prediction time-domain steps. The weighting coefficients represent the state error. This indicates the control increment. Indicates the first The predicted position of the overall center of gravity of the step and the reference target position. Indicates the first ZMP stability margin of the step, Indicates the first The synchronization error between the two bipedal robots during walking. This indicates the increment of the control input.

[0019] The second aspect of the present invention proposes: A control system for the connection transition period of a bipedal robot based on MPC-CPG fusion, wherein the system includes: The calculation module is used to collect IMU data, joint state data and six-dimensional sensor data of the two bipedal robots during the transition phase in response to the corresponding connection signal after the connection between the two bipedal robots is detected, so as to calculate the corresponding overall center of gravity position. The update module is used to initialize the four Hopf oscillator CPG networks and start the corresponding time-varying weight function to generate the initial state of the CPG of the four feet and the continuous foot reference trajectory. In each control cycle, the ZMP position, overall center position and CPG state of the two bipedal robots are updated to construct the corresponding MPC prediction model. The generation module is used to solve the objective function through the MPC prediction model to output the first optimal CPG parameters of the two bipedal robots, and update the first optimal CPG parameters to the Hopf oscillator in real time to generate the corresponding joint commands. The control module is used to send the joint commands to the two bipedal robots, and after determining that the two bipedal robots meet the transition end conditions, it exits the transition phase and synchronously switches to steady-state quadruped CPG planning to complete the corresponding control.

[0020] Furthermore, the expression for the algorithm that calculates the corresponding overall center of gravity position is as follows:

[0021] in, This represents the coordinates of the overall center of gravity of the quadrupedal system formed after connection. The masses of the two individual bipedal robots are respectively. These are the coordinates of their respective centroid positions at the current moment.

[0022] Furthermore, the update module is specifically used for: After entering the transition phase, four mutually coupled improved Hopf oscillator networks are used as rhythmic trajectory generators for the four feet, corresponding to the left front leg, front right leg, rear left leg, and rear right leg of the two bipedal robots, respectively. The leg mapping rules are as follows: the left leg of the first bipedal robot is mapped to the front left leg of a quadruped, and the right leg is mapped to the front right leg of a quadruped; the left leg of the second bipedal robot is mapped to the rear left leg of a quadruped, and the right leg is mapped to the rear right leg of a quadruped.

[0023] Furthermore, the update module is specifically used for: In each control cycle of the transition phase, a time-varying weight function is used as a fusion mechanism to smoothly map the original gait state of the two bipedal robots to the initial CPG state of the four feet in real time. A continuous foot reference trajectory is generated based on the mapped initial state of the CPG.

[0024] Furthermore, the expression for the algorithm that smoothly maps the original gait states of the two bipedal robots to the initial CPG states of the four feet in real time is as follows:

[0025] in, quad,i (0) represents the initial phase used to initialize the i-th leg of the quadrupedal CPG after fusion mapping, and w(t) represents the time-varying weight function. biped,orig This indicates the original bipedal gait phase state of the two robots. quad,ref The reference phase representing the target's quadrupedal gait; The expression for the time-varying weighting function is:

[0026] in, Represents the time-varying weighting function. This represents the current time variable after entering the transition phase. Indicates the attenuation coefficient; The expression for the algorithm that generates a continuous foot reference trajectory based on the mapped initial state of the CPG is as follows:

[0027] in, Indicates the reference trajectory of the foot. Indicates the position of the overall center of gravity. Indicates the output signal of the oscillator. Indicates amplitude. Indicates the offset. This indicates the height offset.

[0028] Furthermore, the expression for the MPC prediction model is:

[0029] in, These represent the system state transition matrix and the control input matrix, respectively. Represents the control input vector. The external disturbance term received by the system is represented by x(k), the state vector is represented by x(k+1), and the MPC prediction model is represented by x(k+1).

[0030] Furthermore, the expression for the objective function is:

[0031] in, This represents the objective function for MPC rolling optimization. Indicates the number of prediction time-domain steps. The weighting coefficients represent the state error. This indicates the control increment. Indicates the first The predicted position of the overall center of gravity of the step and the reference target position. Indicates the first ZMP stability margin of the step, Indicates the first The synchronization error between the two bipedal robots during walking. This indicates the increment of the control input.

[0032] The third aspect of the present invention proposes: A computer includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the control method for the connection transition period of a bipedal robot based on MPC-CPG fusion as described above.

[0033] The fourth aspect of the present invention proposes: A readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the control method for the connection transition period of a bipedal robot based on MPC-CPG fusion as described above.

[0034] The beneficial effects of this invention are as follows: This solution effectively solves the core problem of dynamic balance control during the transition period of robot gait switching in the prior art, significantly improving the robot's adaptability to complex terrain. By synchronously acquiring IMU, joint state, and six-dimensional force data of the foot during the transition phase after the dual-machine connection, the overall center of gravity position of the dual machines can be accurately calculated, providing a reliable state basis for balance control; based on a four-channel Hopf oscillator CPG network with a time-varying weight function, a continuous and smooth foot reference trajectory can be generated, avoiding the impact of trajectory abrupt changes caused by gait switching; combined with the real-time update of ZMP position, overall center of gravity, and CPG state within the control cycle, an MPC prediction model is constructed, which can solve for the optimal CPG parameters in a closed loop and update the oscillator state in real time, ensuring the dynamic balance stability during the transition period; finally, a smooth and disturbance-free switching of the dual machines from the transition phase to the steady-state quadruped CPG planning is achieved, greatly improving the walking stability and environmental adaptability of the dual-machine linkage system in complex terrain.

[0035] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0036] Figure 1 A flowchart of a control method for the connection transition period of a bipedal robot based on MPC-CPG fusion provided in the first embodiment of the present invention; Figure 2 The structural block diagram of the control system for the connection transition period of a bipedal robot based on MPC-CPG fusion provided in the third embodiment of the present invention.

[0037] The following detailed description, in conjunction with the accompanying drawings, will further illustrate the present invention. Detailed Implementation

[0038] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Several embodiments of the invention are illustrated in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0039] It should be noted that when a component is said to be "fixed to" another component, it can be directly on the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0040] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0041] Please see Figure 1 The figure shows the control method for the transition period of bipedal robot connection based on MPC-CPG fusion provided in the first embodiment of the present invention. The control method for the transition period of bipedal robot connection based on MPC-CPG fusion provided in this embodiment can ultimately realize the smooth and uninterrupted switching of the two robots from the transition stage to the steady-state quadruped CPG planning, which greatly improves the walking stability and environmental adaptability of the two-robot linkage system in complex terrain.

[0042] Specifically, this embodiment provides: A control method for the connection transition period of a bipedal robot based on MPC-CPG fusion, wherein the method includes: Step S10: When the connection between the two bipedal robots is detected, in response to the corresponding connection signal, the IMU data, joint status data and six-dimensional sensor data of the two bipedal robots are collected during the transition phase to calculate the corresponding overall center of gravity position. It should be noted that the transition control is triggered by a signal indicating that the connection between the two bipedal robots is complete. This signal usually comes from the positioning sensor of the connection mechanism between the two robots, the feedback signal of the locking mechanism, or the connection completion determination result of the fusion of visual positioning and force perception. Only after confirming that the mechanical connection is completely locked and the two robots form a stable coupled whole will the control logic of the transition phase be triggered, thus completely avoiding safety risks such as mechanism jamming and robot tipping caused by erroneous gait switching when the connection is not completed. After the transition phase begins, the control system will synchronously collect full state data from both bipedal robots. This includes: IMU data installed on the torsos of both robots, covering the three-dimensional attitude angles, three-axis accelerations, and three-axis angular velocities of the torsos, used to accurately obtain the real-time pose and motion state of the two robot torsos, which is the core benchmark for calculating the overall pose of the assembly; state data of all joints of both robots, including the real-time angles, angular velocities, and output torques of each joint such as the hip, knee, and ankle, combined with the robot's DH kinematic parameters, the real-time spatial position of each link and each foot can be obtained through forward kinematics calculation, providing complete link pose data for overall center of gravity calculation; and data from six-dimensional force / torque sensors installed on the four feet, covering the three-dimensional contact force and three-dimensional torque of each foot in contact with the ground, used on the one hand to accurately determine the ground contact state and the switching between support phase and swing phase of each foot, and on the other hand to obtain the support reaction force of the foot, providing real force feedback data for subsequent ZMP calculations and dynamic model construction.

[0043] After collecting all state data, the control system calculates the overall center of gravity position of the two robots after they are connected, based on the principle of rigid body combination center of gravity calculation and the mass distribution parameters of the links and the real-time link poses of the two robots. Unlike the independent center of gravity calculation of a single bipedal robot, the combined body after connection is a new whole formed by the coupling of two independent rigid bodies through the connecting mechanism. All links, torsos, and connecting mechanisms of the two robots must be included in the center of gravity calculation range in order to obtain an accurate overall center of gravity position. The stability of the overall center of gravity is one of the core objectives of the entire transition gait control and also the core state variable for the subsequent MPC prediction model construction.

[0044] Step S20: Initialize four Hopf oscillator CPG networks and start the corresponding time-varying weight function to generate the initial state of the CPG at the four feet and the continuous foot reference trajectory. In each control cycle, update the ZMP position, overall center position and CPG state of the two bipedal robots to construct the corresponding MPC prediction model. It should be noted that gait generation for a single bipedal robot typically uses a CPG network consisting of two Hopf oscillators, corresponding to the left and right feet respectively. When two bipedal robots are connected, a quadruped configuration with four feet is formed. Therefore, it is necessary to initialize four mutually coupled Hopf oscillator CPG networks. Each Hopf oscillator corresponds to one foot of the assembly. By setting the phase difference between the oscillators, the basic timing sequence of quadruped gait can be generated, such as diagonal trotting gait, shuffling gait, and other stable gaits suitable for the transition period. As a nonlinear oscillator, the core advantage of the Hopf oscillator is that it can generate continuous, smooth, and periodic rhythmic signals, and its parameters are easy to adjust. By modifying the amplitude, frequency, phase, and bias of the oscillator, the stride length, stride height, gait period, and foot placement can be quickly adjusted, perfectly adapting to the dynamic adjustment requirements of gait during the transition period. Simultaneously, this step activates the corresponding time-varying weight function. The core function of this function is to achieve a smooth transition of gait and avoid the abrupt shock from bipedal gait to quadrupedal gait. In the early stage of the transition phase, the time-varying weight retains some features of the original bipedal gait of the two robots. As the transition phase progresses, the weight will gradually and smoothly switch to the parameters of quadrupedal coordinated gait, allowing the gait of the four feet to gradually transition from the original independent bipedal rhythm of the two robots to a coordinated quadrupedal rhythm, completely eliminating the joint shock and center of gravity fluctuation caused by the gait abrupt change.

[0045] Within each control cycle, the control system synchronously updates three core state variables: First, the real-time ZMP position of the two robots. ZMP (Zero Moment Point) is the core criterion for the dynamic stability of legged robots. Only when the ZMP always falls within the support polygon formed by the four legs will the assembly not tip over. The support polygon after connection changes from a line segment shape of two legs to a quadrilateral formed by four legs, greatly expanding the support domain, but precise real-time calculation is still required to ensure stability. Second, the overall center of gravity position of the assembly, which is updated in real time based on the latest acquired state data to ensure the accuracy of the center of gravity position. Third, the real-time state of the CPG network, including the real-time values ​​of parameters such as the phase, amplitude, and frequency of the four Hopf oscillators. Based on these real-time updated state variables, the control system constructs a corresponding MPC prediction model. This model is a discretized linear time-varying model that includes the dynamic characteristics of the assembly, the mapping relationship between CPG parameters and foot trajectory, ZMP stability constraints, and joint motion constraints. It can accurately predict the trends of changes in the center of gravity, ZMP position, foot trajectory, and joint state of the assembly under different CPG parameter adjustments in the prediction time domain, providing an accurate prediction basis for subsequent rolling optimization solutions.

[0046] Step S30: Solve the objective function through the MPC prediction model to output the first optimal CPG parameters of the two bipedal robots, and update the first optimal CPG parameters to the Hopf oscillator in real time to generate the corresponding joint commands; It should be noted that the control system, based on the constructed MPC prediction model, solves a constrained quadratic programming problem within the preset prediction and control time domains, using the set objective function as the optimization objective and ZMP stability constraints, joint angle / angular velocity / torque constraints, foot movement range constraints, and center of gravity fluctuation constraints as boundary conditions. The objective function is the core of the entire optimization problem, typically consisting of multiple weighted optimization sub-terms. These include: 1) a ZMP tracking error term, minimizing the deviation between the actual and desired ZMP positions to ensure the ZMP always falls within the supporting polygon, guaranteeing the dynamic stability of the assembly; 2) a center of gravity tracking error term, minimizing the deviation between the actual overall center of gravity and the desired center of gravity trajectory to ensure the stability of the assembly's center of gravity and avoid large fluctuations; 3) a foot trajectory tracking error term, minimizing the deviation between the actual foot trajectory and the reference trajectory generated by the CPG to ensure the rhythmicity of the gait; 4) a control quantity smoothing term, minimizing the rate of change of CPG parameters to avoid gait shocks and joint vibrations caused by sudden parameter changes; and 5) a joint energy consumption optimization term, minimizing fluctuations in the joint's output torque to reduce energy consumption during the transition period.

[0047] By setting reasonable weights for these sub-items, an optimal balance can be achieved between stability, smoothness, and energy consumption. The core feature of MPC is rolling optimization. Within each control cycle, the solver obtains multiple sets of optimal CPG parameters in the control time domain, but only outputs the first optimal CPG parameters for the first control cycle, including core parameters such as amplitude, frequency, phase, and bias for the four Hopf oscillators. This rolling optimization mode allows the control system to re-optimize based on the latest state data in each control cycle, significantly improving the control's anti-disturbance capability and real-time performance, and enabling rapid response to model uncertainties and external disturbances in the connected assembly. After obtaining the optimal CPG parameters, the control system updates these parameters to the four Hopf oscillators in real time. Based on the updated parameters, the oscillators output smooth periodic rhythmic signals, mapped to Cartesian space reference trajectories at the four feet. Then, through inverse kinematics calculations of the two robots, the spatial trajectories of the feet are converted into angle, angular velocity, and torque commands for each joint—the final joint control commands—achieving a complete conversion from optimal stability parameters to joint execution commands.

[0048] Step S40: The joint command is sent to the two bipedal robots. After determining that the two bipedal robots meet the transition end condition, the transition phase is exited and the system is switched to steady-state quadruped CPG planning to complete the corresponding control.

[0049] It should be noted that the control system synchronously sends the calculated joint commands to the joint controllers of the two bipedal robots via a real-time industrial Ethernet bus. The joint controllers drive the corresponding joints to move in real time according to the commands, executing the four-legged coordinated gait during the transition phase. At the same time, the control system continuously collects the robot's status feedback data in each control cycle, forming a complete closed-loop control of "collection-modeling-optimization-execution-feedback" to ensure the accuracy of gait execution and the stability of the combined robot. During gait execution, the control system continuously judges whether the assembly meets the transition termination conditions. The transition termination conditions are the core basis for determining whether the assembly has smoothly transitioned to a stable quadrupedal gait, and usually include multiple quantitative judgment criteria: First, the fluctuation amplitude of the overall center of gravity is less than the preset stability threshold for several consecutive gait cycles, indicating that the center of gravity of the assembly has been completely stabilized without large fluctuations; Second, the ZMP position is stably located within the safe area of ​​the supporting polygon for several consecutive gait cycles without the risk of exceeding the boundary, indicating that the dynamic stability of the assembly fully meets the requirements; Third, the ground contact timing, gait cycle, and phase difference of the four feet remain stable for several consecutive cycles, indicating that the quadrupedal coordinated gait has been completely converged without rhythmic disorder; Fourth, the fluctuation of joint torque and angular velocity is less than the preset threshold, indicating that the gait is without impact or shaking, and the execution process is smooth.

[0050] The control system will only determine the end of the transition phase when all the judgment conditions are met consecutively, avoiding instability caused by premature switching. After the transition is determined to be over, the control system will exit the MPC-CPG fusion control mode of the transition phase and simultaneously switch to the steady-state quadruped CPG planning mode. This is because the core objective of the transition phase is to achieve a smooth gait transition, which requires high-frequency rolling optimization of MPC to ensure stability. However, after entering the steady state, the quadruped gait of the combined robot is completely stable, and the control requirements can be met by using open-loop steady-state quadruped CPG planning. This can significantly reduce the computational load of the control system, improve the real-time performance and endurance of the control, and also adapt to different walking speeds and terrain environments through the adjustment of steady-state CPG parameters, so as to complete the long-term collaborative operation task after the two robots are connected. Ultimately, the full-process control goal of transitioning from bipedal independent walking to smooth transition after connection, and then to steady-state quadrupedal collaborative walking is achieved.

[0051] Second Embodiment Furthermore, the expression for the algorithm that calculates the corresponding overall center of gravity position is as follows:

[0052] in, This represents the coordinates of the overall center of gravity of the quadrupedal system formed after connection. The masses of the two individual bipedal robots are respectively. These are the coordinates of their respective centroid positions at the current moment.

[0053] Furthermore, the step of initializing the four Hopf oscillator CPG network includes: After entering the transition phase, four mutually coupled improved Hopf oscillator networks are used as rhythmic trajectory generators for the four feet, corresponding to the left front leg, front right leg, rear left leg, and rear right leg of the two bipedal robots, respectively. The leg mapping rules are as follows: the left leg of the first bipedal robot is mapped to the front left leg of a quadruped, and the right leg is mapped to the front right leg of a quadruped; the left leg of the second bipedal robot is mapped to the rear left leg of a quadruped, and the right leg is mapped to the rear right leg of a quadruped.

[0054] Furthermore, the step of activating the corresponding time-varying weighting function to generate the initial CPG states of the four foot ends and the continuous foot reference trajectory includes: In each control cycle of the transition phase, a time-varying weight function is used as a fusion mechanism to smoothly map the original gait state of the two bipedal robots to the initial CPG state of the four feet in real time. A continuous foot reference trajectory is generated based on the mapped initial state of the CPG.

[0055] Furthermore, the expression for the algorithm that smoothly maps the original gait states of the two bipedal robots to the initial CPG states of the four feet in real time is as follows:

[0056] in, quad,i (0) represents the initial phase used to initialize the i-th leg of the quadrupedal CPG after fusion mapping, and w(t) represents the time-varying weight function. biped,orig This indicates the original bipedal gait phase state of the two robots. quad,ref The reference phase representing the target's quadrupedal gait; The expression for the time-varying weighting function is:

[0057] in, Represents the time-varying weighting function. This represents the current time variable after entering the transition phase. Indicates the attenuation coefficient; The expression for the algorithm that generates a continuous foot reference trajectory based on the mapped initial state of the CPG is as follows:

[0058] in, Indicates the reference trajectory of the foot. Indicates the position of the overall center of gravity. Indicates the output signal of the oscillator. Indicates amplitude. Indicates the offset. This indicates the height offset.

[0059] Furthermore, the expression for the MPC prediction model is:

[0060] in, These represent the system state transition matrix and the control input matrix, respectively. Represents the control input vector. The external disturbance term received by the system is represented by x(k), the state vector is represented by x(k+1), and the MPC prediction model is represented by x(k+1).

[0061] Furthermore, the expression for the objective function is:

[0062] in, This represents the objective function for MPC rolling optimization. Indicates the number of prediction time-domain steps. The weighting coefficients represent the state error. This indicates the control increment. Indicates the first The predicted position of the overall center of gravity of the step and the reference target position. Indicates the first ZMP stability margin of the step, Indicates the first The synchronization error between the two bipedal robots during walking. This indicates the increment of the control input.

[0063] In addition, it should be noted in this embodiment that the control method for the connection transition period of a bipedal robot based on MPC-CPG fusion provided in this embodiment is specifically implemented by the following: This invention provides a dynamic balance control method based on MPC-CPG fusion for the transition period after two bipedal robots are connected, which includes the following four levels: Basic generation layer A dedicated transition phase (duration) after the connection completion signal is triggered. (s, which can be adaptively adjusted according to actual dynamic parameters) First, the basic quadrupedal rhythm trajectory is generated by the improved Hopf oscillator CPG network, realizing the accurate mapping from the legs of the two bipedal robots to the front / back legs of the quadrupeds; Transition fusion layer Includes time-varying weight function .

[0064] : A time-varying weighting function used to smoothly map gait states during the transition period. This value changes over time from... Smooth transition to .

[0065] : The current time variable after entering the dedicated transition phase.

[0066] The attenuation coefficient (or convergence rate parameter) determines the speed of the fusion transition. This embodiment recommends an optimal value of [value missing]. .

[0067] MPC Optimization Layer MPC controller with CPG key parameter vector (Phase difference, amplitude, offset) are the control variables. The rolling optimization is performed within the finite prediction time domain. The objective function includes minimizing the overall center of gravity trajectory deviation, maximizing the ZMP margin, minimizing the synchronization error of the two machines, and joint torque smoothing constraints. Observation and Judgment Layer A connection disturbance observer is constructed by using six-dimensional force / torque feedback at the foot end to compensate for the center of gravity shift disturbance at the moment of connection in real time. Finally, only the first step of the optimal CPG parameters output by MPC are sent to the two robot joint controllers to achieve a transition with no posture disturbance and continuous torque, until ZMP stabilizes and the quadrupedal gait parameters converge, and then smoothly switch to steady-state quadrupedal CPG planning.

[0068] The innovation of this invention lies in the following four process steps, each of which addresses a specific dynamic change problem during the transition period after two bipedal robots are connected, providing a complete technical solution and specific implementation details that differ from existing single-robot control technologies: Step 1: Improve the CPG base generation layer Upon connection completion signal triggering, a dedicated transition phase is immediately initiated. Four mutually coupled improved Hopf oscillator networks are used as the quadrupedal rhythmic trajectory generator, corresponding to the front left leg, front right leg, rear left leg, and rear right leg of the quadrupedal structure, respectively. The leg mapping rules are as follows: for the first bipedal robot (Robot1), the left leg is mapped to the front left leg of the quadruped, and the right leg is mapped to the front right leg; for the second bipedal robot (Robot2), the left leg is mapped to the rear left leg of the quadruped, and the right leg is mapped to the rear right leg. For the trot gait, the coupling matrix... Designed between the front and back legs The fixed phase difference. The oscillator parameters are set to... Initial frequency .

[0069] as well as;

[0070] : No. The natural angular frequency of the Hopf oscillator.

[0071] Reference value for the basic step frequency of a quadruped robot (unit: Hz).

[0072] : No. The target value for the amplitude convergence of the oscillator.

[0073] : No. The expected amplitude of each leg. Within this framework, It is the control quantity calculated and issued by the MPC optimization layer.

[0074] Step 2: Transition and Blending Layer During each control cycle of the transition phase, a time-varying weighting function is used. As a fusion mechanism, the original gait states of the two bipedal robots are smoothly mapped to the initial state of the quadrupedal CPG in real time. The specific calculation formula is as follows:

[0075] quad,i(0): After fusion mapping, the initial phase of the i-th leg of the quadruped CPG is used to initialize the initial phase.

[0076] w(t): Time-varying weight function.

[0077] biped,orig: The original bipedal gait phase state of the two robots.

[0078] quad,ref: The reference phase of the target quadrupedal gait (such as trot gait).

[0079] Simultaneously, a continuous foot reference trajectory is generated based on the mapped CPG state:

[0080] (Foot reference trajectory): The first The desired spatial position coordinates of a leg relative to the global coordinate system.

[0081] (Overall center of gravity position): The unified center of gravity of the quadruped system is calculated by the mass weights of the two bipedal robots.

[0082] (Oscillator output signal): The normalized periodic signal generated by the improved Hopf oscillator determines the rhythm and phase of the gait.

[0083] (Amplitude): The optimal amplitude parameter issued by the MPC optimization layer directly determines the size of the step.

[0084] (Offset): An offset term belonging to the CPG parameter vector. It is used to adjust the center position of the foot's landing point.

[0085] (Height offset): Determines the vertical height of the foot lift or support height, ensuring sufficient obstacle-crossing space in complex terrain.

[0086] Physical meaning: It combines "center of gravity position" with "stepping motion of the foot relative to the center of gravity." Real-time adjustments are made via MPC. and It can dynamically widen or narrow the supporting polygon, thereby quickly stabilizing its shape during the transition period.

[0087] Step 3: MPC Optimization Layer Build an MPC controller specifically for the transition period, and vectorize the CPG key parameters. As the sole control variable, in the prediction time domain ( s) In-process scrolling optimization. The state vector is defined as:

[0088] The system in discrete time steps The state vector.

[0089] The positional components of the overall center of gravity of the assembly along the X and Y axes.

[0090] : The velocity components of the overall center of gravity of the combined body in the X and Y axis directions.

[0091] : The actual position of the current zero torque point (ZMP) in the X and Y axis directions.

[0092] The MPC prediction model is:

[0093] : System state transition matrix and control input matrix.

[0094] : Control input vector (i.e., CPG key parameter vector) ).

[0095] External disturbances experienced by the system (such as transient shocks at the moment of connection).

[0096] The objective function to be optimized is:

[0097] : The objective function for MPC rolling optimization.

[0098] The number of time-domain steps for prediction is set in the paper as follows: .

[0099] : Weighting coefficients for state errors. These correspond to the overall center of gravity trajectory deviation, ZMP margin, and synchronization error between the two machines, respectively.

[0100] : Penalty weight coefficient matrix for control increment (joint torque smoothing constraint).

[0101] : No. The overall center of gravity prediction position and reference target position of the step.

[0102] : No. The ZMP stability margin of the step (i.e., the distance of ZMP from the edge of the supporting polygon).

[0103] : No. The synchronization error between the two bipedal robots during walking.

[0104] Torque smoothing can be achieved by controlling the increment of the input (for the rate of change of stress torque) and limiting its magnitude.

[0105] (Recommendation weight) , , , The OSQP quadratic programming solver is used, and only the optimal CPG parameters from the first step are distributed.

[0106] Step 4: Observation and Judgment Layer A connection disturbance observer is constructed in real time based on feedback from four foot-end six-dimensional force / torque sensors. .

[0107] The centroid shift disturbance value estimated by the observer at the instant of connection.

[0108] : Observer gain matrix or coefficients.

[0109] The contact force in the Z-direction (vertical direction) is measured in real time by a six-dimensional force sensor at the foot.

[0110] Ideal Z-direction contact force predicted based on kinematic and dynamic models.

[0111] The center of gravity shift disturbance at the moment of connection is estimated and feedforward compensated. Simultaneously, the overall center of gravity position is calculated in real time:

[0112] : The coordinates of the center of gravity of the four-legged system formed after connection.

[0113] : These are the masses of two individual bipedal robots, Robot1 and Robot2, respectively.

[0114] : These are the coordinates of the centroid positions of Robot1 and Robot2 at the current moment.

[0115] Transition termination condition: When ZMP margin is reached. m and When m continues for 8 sampling cycles, it automatically exits the transition phase and switches to steady-state quadrupedal CPG programming.

[0116] Technical effect Attitude fluctuations were significantly reduced during the transition period (roll / pitch fluctuations) ZMP always remains within the safety margin of the quadrupedal supported polygon. m), completely eliminating the risk of overturning; The transition time is stably controlled within 2 seconds, requiring no manual intervention and without sudden torque changes, achieving true "hot switching"; Switching success rate The overall stability of the combined unit is superior to that of a single bipedal unit, allowing it to immediately enter a walking mode for complex terrain.

[0117] Please see Figure 2 The third embodiment of the present invention provides: A control system for the connection transition period of a bipedal robot based on MPC-CPG fusion, wherein the system includes: The calculation module is used to collect IMU data, joint state data and six-dimensional sensor data of the two bipedal robots during the transition phase in response to the corresponding connection signal after the connection between the two bipedal robots is detected, so as to calculate the corresponding overall center of gravity position. The update module is used to initialize the four Hopf oscillator CPG networks and start the corresponding time-varying weight function to generate the initial state of the CPG of the four feet and the continuous foot reference trajectory. In each control cycle, the ZMP position, overall center position and CPG state of the two bipedal robots are updated to construct the corresponding MPC prediction model. The generation module is used to solve the objective function through the MPC prediction model to output the first optimal CPG parameters of the two bipedal robots, and update the first optimal CPG parameters to the Hopf oscillator in real time to generate the corresponding joint commands. The control module is used to send the joint commands to the two bipedal robots, and after determining that the two bipedal robots meet the transition end conditions, it exits the transition phase and synchronously switches to steady-state quadruped CPG planning to complete the corresponding control.

[0118] Furthermore, the expression for the algorithm that calculates the corresponding overall center of gravity position is as follows:

[0119] in, This represents the coordinates of the overall center of gravity of the quadrupedal system formed after connection. The masses of the two individual bipedal robots are respectively. These are the coordinates of their respective centroid positions at the current moment.

[0120] Furthermore, the update module is specifically used for: After entering the transition phase, four mutually coupled improved Hopf oscillator networks are used as rhythmic trajectory generators for the four feet, corresponding to the left front leg, front right leg, rear left leg, and rear right leg of the two bipedal robots, respectively. The leg mapping rules are as follows: the left leg of the first bipedal robot is mapped to the front left leg of a quadruped, and the right leg is mapped to the front right leg of a quadruped; the left leg of the second bipedal robot is mapped to the rear left leg of a quadruped, and the right leg is mapped to the rear right leg of a quadruped.

[0121] Furthermore, the update module is specifically used for: In each control cycle of the transition phase, a time-varying weight function is used as a fusion mechanism to smoothly map the original gait state of the two bipedal robots to the initial CPG state of the four feet in real time. A continuous foot reference trajectory is generated based on the mapped initial state of the CPG.

[0122] Furthermore, the expression for the algorithm that smoothly maps the original gait states of the two bipedal robots to the initial CPG states of the four feet in real time is as follows:

[0123] in, quad,i (0) represents the initial phase used to initialize the i-th leg of the quadrupedal CPG after fusion mapping, and w(t) represents the time-varying weight function. biped,orig This indicates the original bipedal gait phase state of the two robots. quad,ref The reference phase representing the target's quadrupedal gait; The expression for the time-varying weighting function is:

[0124] in, Represents the time-varying weighting function. This represents the current time variable after entering the transition phase. Indicates the attenuation coefficient; The expression for the algorithm that generates a continuous foot reference trajectory based on the mapped initial state of the CPG is as follows:

[0125] in, Indicates the reference trajectory of the foot. Indicates the position of the overall center of gravity. Indicates the output signal of the oscillator. Indicates amplitude. Indicates the offset. This indicates the height offset.

[0126] Furthermore, the expression for the MPC prediction model is:

[0127] in, These represent the system state transition matrix and the control input matrix, respectively. Represents the control input vector. The external disturbance term received by the system is represented by x(k), the state vector is represented by x(k+1), and the MPC prediction model is represented by x(k+1).

[0128] Furthermore, the expression for the objective function is:

[0129] in, This represents the objective function for MPC rolling optimization. Indicates the number of prediction time-domain steps. The weighting coefficients represent the state error. This indicates the control increment. Indicates the first The predicted position of the overall center of gravity of the step and the reference target position. Indicates the first ZMP stability margin of the step, Indicates the first The synchronization error between the two bipedal robots during walking. This indicates the increment of the control input.

[0130] The fourth embodiment of the present invention provides a computer, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the control method for the connection transition period of a bipedal robot based on MPC-CPG fusion as described above.

[0131] The fifth embodiment of the present invention provides a readable storage medium storing a computer program thereon, wherein when the program is executed by a processor, it implements the control method for the connection transition period of a bipedal robot based on MPC-CPG fusion as described above.

[0132] In summary, the control method and system for the transition period of bipedal robot connection based on MPC-CPG fusion provided in the above embodiments of the present invention can ultimately achieve a smooth and uninterrupted switch between the two robots from the transition phase to the steady-state quadruped CPG planning, which greatly improves the walking stability and environmental adaptability of the dual-robot linkage system in complex terrain.

[0133] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can reside in the same processor; or the above modules can be located in different processors in any combination.

[0134] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0135] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0136] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0137] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0138] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. A control method for the connection transition period of a bipedal robot based on MPC-CPG fusion, characterized in that, The method includes: Once the connection between the two bipedal robots is detected, in response to the corresponding connection signal, the IMU data, joint status data, and six-dimensional sensor data of the four feet of the two bipedal robots are collected during the transition phase to calculate the corresponding overall center of gravity position. Initialize four Hopf oscillator CPG networks and start the corresponding time-varying weight function to generate the initial state of the CPG at the four feet and the continuous foot reference trajectory. In each control cycle, update the ZMP position, overall center position and CPG state of the two bipedal robots to construct the corresponding MPC prediction model. The objective function is solved by the MPC prediction model to output the first optimal CPG parameters of the two bipedal robots, and the first optimal CPG parameters are updated to the Hopf oscillator in real time to generate the corresponding joint commands. The joint commands are sent to the two bipedal robots. After determining that the two bipedal robots meet the transition end conditions, the transition phase is exited and the system is switched to steady-state quadruped CPG planning to complete the corresponding control. The step of activating the corresponding time-varying weight function to generate the initial CPG states of the four feet and the continuous foot reference trajectory includes: In each control cycle of the transition phase, a time-varying weight function is used as a fusion mechanism to smoothly map the original gait state of the two bipedal robots to the initial CPG state of the four feet in real time. Generate a continuous foot reference trajectory based on the mapped CPG initial state; The algorithm for smoothly mapping the original gait states of the two bipedal robots to the initial CPG states of the four feet in real time is expressed as follows: in, quad,i (0) represents the initial phase used to initialize the i-th leg of the quadrupedal CPG after fusion mapping, and w(t) represents the time-varying weight function. biped,orig This indicates the original bipedal gait phase state of the two robots. quad,ref The reference phase representing the target's quadrupedal gait; The expression for the time-varying weighting function is: in, Represents the time-varying weighting function. This represents the current time variable after entering the transition phase. Indicates the attenuation coefficient; The expression for the algorithm that generates a continuous foot reference trajectory based on the mapped initial state of the CPG is as follows: in, Indicates the reference trajectory of the foot. Indicates the position of the overall center of gravity. Indicates the output signal of the oscillator. Indicates amplitude. Indicates the offset. This indicates the height offset.

2. The control method for the connection transition period of a bipedal robot based on MPC-CPG fusion according to claim 1, characterized in that, The expression for the algorithm that calculates the corresponding overall center of gravity position is: in, This represents the coordinates of the overall center of gravity of the quadrupedal system formed after connection. The masses of the two individual bipedal robots are respectively. These are the coordinates of their respective centroid positions at the current moment.

3. The control method for the connection transition period of a bipedal robot based on MPC-CPG fusion according to claim 1, characterized in that, The steps for initializing the four Hopf oscillator CPG network include: After entering the transition phase, four mutually coupled improved Hopf oscillator networks are used as rhythmic trajectory generators for the four feet, corresponding to the left front leg, front right leg, rear left leg, and rear right leg of the two bipedal robots, respectively. The leg mapping rules are as follows: the left leg of the first bipedal robot is mapped to the front left leg of a quadruped, and the right leg is mapped to the front right leg of a quadruped; the left leg of the second bipedal robot is mapped to the rear left leg of a quadruped, and the right leg is mapped to the rear right leg of a quadruped.

4. The control method for the connection transition period of a bipedal robot based on MPC-CPG fusion according to claim 1, characterized in that, The expression for the MPC prediction model is: in, These represent the system state transition matrix and the control input matrix, respectively. This represents the control input vector. The external disturbance term received by the system is represented by x(k), the state vector is represented by x(k+1), and the MPC prediction model is represented by x(k+1).

5. The control method for the connection transition period of a bipedal robot based on MPC-CPG fusion according to claim 1, characterized in that, The expression for the objective function is: in, This represents the objective function for MPC rolling optimization. Indicates the number of prediction time-domain steps. The weighting coefficients represent the state error. This indicates the control increment. Indicates the first The predicted position of the overall center of gravity of the step and the reference target position. Indicates the first ZMP stability margin of the step, Indicates the first The synchronization error between the two bipedal robots during walking. This indicates the increment of the control input.

6. A control system for the connection transition period of a bipedal robot based on MPC-CPG fusion, characterized in that, For implementing the control method for the connection transition period of a bipedal robot based on MPC-CPG fusion as described in any one of claims 1 to 5, the system comprises: The calculation module is used to collect IMU data, joint state data and six-dimensional sensor data of the two bipedal robots during the transition phase in response to the corresponding connection signal after the connection between the two bipedal robots is detected, so as to calculate the corresponding overall center of gravity position. The update module is used to initialize the four Hopf oscillator CPG networks and start the corresponding time-varying weight function to generate the initial state of the CPG of the four feet and the continuous foot reference trajectory. In each control cycle, the ZMP position, overall center position and CPG state of the two bipedal robots are updated to construct the corresponding MPC prediction model. The generation module is used to solve the objective function through the MPC prediction model to output the first optimal CPG parameters of the two bipedal robots, and update the first optimal CPG parameters to the Hopf oscillator in real time to generate the corresponding joint commands. The control module is used to send the joint commands to the two bipedal robots, and after determining that the two bipedal robots meet the transition end conditions, it exits the transition phase and synchronously switches to steady-state quadruped CPG planning to complete the corresponding control.

7. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the control method for the connection transition period of a bipedal robot based on MPC-CPG fusion as described in any one of claims 1 to 5.

8. A readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the control method for the connection transition period of a bipedal robot based on MPC-CPG fusion as described in any one of claims 1 to 5.

Images