A robot gait driving control method and a multi-legged robot driving system

Abstract

The application provides a robot gait driving control method and a multi-legged robot driving system, and relates to the technical field of robot control. The application realizes effective separation between high-level intelligent decision and high real-time joint servo control by introducing a distributed control architecture, reduces the single-point failure risk of the multi-legged robot driving system, improves the reliability of the driving system, realizes gait dynamic adjustment function following terrain changes in an unstructured environment through deep integration of a multi-modal perception mechanism and an adaptive control algorithm in the high-level intelligent decision process, and effectively improves joint dynamic response accuracy by introducing a force-position hybrid double closed-loop control mechanism in the joint servo control process, so as to ensure smooth and fluent adaptive gait motion effect of the multi-legged robot in a complex unstructured environment.

Description

Technical Field

[0001] This application relates to the field of robot control technology, and more specifically, to a robot gait drive control method and a multi-legged robot drive system. Background Technology

[0002] With the continuous development of science and technology, robotics has received widespread attention from various industries (such as home services, science education, emergency rescue, etc.) due to its great research and application value. Among them, multi-legged robots built using multi-degree-of-freedom leg structures (such as spider robots, dinosaur bionic robots, etc.) are an important research direction in current robotics technology.

[0003] In practical applications of multi-legged robots, their operating environments often involve complex, unstructured terrain. This necessitates the robot actively adjusting its gait to adapt to changes in the terrain, ensuring smooth and coordinated gait movement. However, it's worth noting that existing multi-legged robot drive systems primarily employ a centralized control architecture (using a single main controller to handle all gait calculation tasks, including gait planning, inverse kinematics solving, and joint servo control) and the classic PID (Proportion-Integral-Differential) control algorithm. This approach suffers from poor real-time control performance and a high risk of single-point failures, making it difficult to adapt to the dynamic gait adjustment requirements, high real-time gait control requirements, and high reliability requirements of drive systems in unstructured environments. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a robot gait drive control method and a multi-legged robot drive system. By introducing a distributed control architecture (i.e., a "central decision-making-distributed execution" architecture), it can effectively separate high-level intelligent decision-making from high-real-time joint servo control, reduce the risk of single-point failure in the multi-legged robot drive system, and improve the reliability of the drive system. In the high-level intelligent decision-making process, through the deep integration of multimodal perception mechanisms (including joint state perception, trunk IMU (Inertial Measurement Unit) perception, visual odometry perception, environmental terrain perception, etc.) and adaptive control algorithms, it can realize the function of dynamic gait adjustment following terrain changes in unstructured environments. At the same time, in the joint servo control process, by introducing a force-position hybrid dual closed-loop control mechanism (i.e., force-position hybrid dual-loop PI control operation), it can effectively improve the joint dynamic response accuracy, thereby ensuring the smooth and adaptive gait movement of the multi-legged robot in complex unstructured environments.

[0005] To achieve the above objectives, the technical solutions adopted in the embodiments of this application are as follows: In a first aspect, this application provides a robot gait drive control method applied to a multi-legged robot drive system. The system includes a central controller and multiple joint controllers that are communicatively connected to each other, wherein each joint controller is used to individually control a leg joint of the multi-legged robot. The method includes: The central controller acquires the actual joint state data uploaded by each of the multiple joint controllers in the current control cycle, and acquires the torso IMU data, visual odometry data and terrain perception data of the multi-legged robot in the current control cycle. The central controller determines the global motion state data of the multi-legged robot for the desired gait trajectory in the current control cycle based on the acquired actual joint state data, trunk IMU data, and visual odometry data. The central controller calls the adaptive control engine to solve for the leg joint motion commands based on the terrain perception data and the global motion state data, and the expected gait trajectory, so as to obtain the expected joint motion commands of each leg joint in the next control cycle. Each joint controller receives the desired joint motion command for the corresponding leg joint from the central controller, and performs force-position hybrid dual-loop PI control on the leg joint according to the desired joint motion command in the next control cycle to drive the leg joint to move.

[0006] In an optional implementation, the global motion state data includes the robot state data of the multi-legged robot and the global tracking error data of the multi-legged robot for the desired gait trajectory. The adaptive control engine includes a gait planning and optimization network, a model prediction control module, and a sliding mode supervision module. The central controller calls the adaptive control engine to solve for the leg joint motion commands for the desired gait trajectory based on the terrain perception data and the global motion state data, obtaining the desired joint motion commands for each of the leg joints in the next control cycle. This includes the following steps: The gait planning and optimization network is invoked to perform gait planning optimization inference based on the robot state data, the terrain perception data, and the desired gait trajectory, to obtain the desired gait parameter compensation amount, the model parameter optimization amount of the model prediction control module, and the sliding mode gain fine-tuning amount of the sliding mode supervision module; The model prediction control module is invoked to perform joint motion prediction for all leg joints based on the global motion state data according to the model parameter tuning amount, so as to obtain the theoretical feedforward torque, theoretical joint angle and theoretical single-leg gait phase of each leg joint in the next control cycle. The sliding mode supervision module is invoked to perform chatter suppression prediction for each of the leg joints based on the global tracking error data according to the sliding mode gain fine-tuning amount, so as to obtain the theoretical chatter suppression compensation torque of each of the leg joints in the next control cycle. Based on the expected gait parameter compensation amount and the expected gait trajectory, the theoretical feedforward torque, theoretical joint angle, theoretical single-leg gait phase and theoretical flutter suppression compensation torque of each of the leg joints in the next control cycle are integrated to obtain the expected joint movement commands of each of the leg joints in the next control cycle that satisfy the joint limit movement constraints.

[0007] In an optional implementation, the desired joint motion command for each leg joint in the next control cycle includes the desired feedforward torque, desired joint angle, desired single-leg gait phase, desired flutter suppression compensation torque, and desired joint angular velocity boundary for the corresponding leg joint. Each control cycle covers multiple outer loop cycles, and each outer loop cycle covers multiple inner loop cycles. The step of each joint controller performing force-position hybrid dual-loop PI control on the corresponding leg joint according to the corresponding desired joint motion command in the next control cycle includes: The system sequentially traverses multiple outer loop cycles covered by the next control cycle. For each traversed outer loop cycle, based on the actual joint angle, actual joint angular velocity, actual leg ground contact probability, and actual foot three-dimensional force of the corresponding leg joint in that outer loop cycle, force-position coordination PI control is performed based on the desired joint angle, the desired single-leg gait phase, the desired flutter suppression compensation torque, and the desired joint angular velocity boundary to obtain the target joint torque of the leg joint in that outer loop cycle. The outer loop cycle covers multiple inner loop cycles, and for each inner loop cycle, based on the target joint torque of the leg joint in the outer loop cycle, torque servo PI control is performed based on the expected feedforward torque, the expected single-leg gait phase, and the actual joint torque of the leg joint in the corresponding inner loop cycle to obtain the expected motor current of the leg joint in the inner loop cycle. The joint motor that drives the leg joint operates according to the corresponding desired motor current during the inner loop cycle.

[0008] In an optional implementation, for each traversed outer loop cycle, the step of performing force-position coordination PI control based on the expected joint angle, the expected single-leg gait phase, the expected flutter suppression compensation torque, and the expected joint angular velocity boundary to obtain the target joint torque of the leg joint in that outer loop cycle, according to the actual joint angle, actual joint angular velocity, actual leg ground contact probability, and actual foot three-dimensional force of the corresponding leg physical joint in that outer loop cycle, includes: Calculate the joint angle admittance offset of the leg joint in the outer circumference period that matches the desired single-leg gait based on the actual three-dimensional force at the foot end. Calculate the joint angle error between the target joint angle and the actual joint angle in the outer loop cycle, and perform force-position outer loop PI control based on the joint angle error according to the expected single-leg gait phase and the actual leg ground contact probability to obtain the reference joint torque of the leg entity joint in the outer loop cycle. The target joint angle in the outer loop cycle is obtained by adding the joint angle admittance offset of the corresponding outer loop cycle and the expected joint angle. Based on the actual joint angular velocity and the desired joint angular velocity boundary, calculate the theoretical damping torque of the leg joint in the outer ring cycle; The reference joint torque, theoretical damping torque, and desired flutter suppression compensation torque of the leg joint in the outer ring cycle are added together to obtain the target joint torque of the leg joint in the outer ring cycle.

[0009] In an optional implementation, the step of performing force-position outer-loop PI control based on the joint angle error according to the desired single-leg gait phase and the actual leg ground contact probability to obtain the reference joint torque of the leg joint in the outer-loop cycle includes: Based on the desired single-leg gait phase and the actual leg contact probability, the start and stop states of the outer loop integral controller of the leg joint in the outer loop cycle are adjusted, and the integral control output of the outer loop integral controller for the joint angle error in the outer loop cycle is determined. The proportional control output of the outer loop proportional controller of the leg joint is calculated for the joint angle error in the outer loop cycle, and the proportional control output and integral control output corresponding to the outer loop cycle are added to obtain the reference joint torque of the leg joint in the outer loop cycle.

[0010] In an optional implementation, the step of adjusting the start / stop state of the outer-loop integral controller of the leg joint in the outer-loop cycle according to the desired single-leg gait phase and the actual leg ground contact probability, and determining the integral control output of the outer-loop integral controller for the joint angle error in the outer-loop cycle, includes: When the desired single-leg gait phase belongs to the support phase and the actual leg contact probability is greater than the first preset probability value, the outer loop integral controller is activated in the outer loop cycle, and the outer loop integral controller is driven to generate the integral control output according to the nominal value of the outer loop integral gain. When the desired single-leg gait phase belongs to the support phase and the actual leg contact probability is between the first preset probability value and the second preset probability value, the outer loop integral controller is activated in the outer loop cycle, and the outer loop integral controller is driven to generate the integral control output according to the outer loop integral gain transition value, wherein the second preset probability value is less than the first preset probability value, and the outer loop integral gain transition value is half of the nominal value of the outer loop integral gain. When the desired single-leg gait phase is a swing phase and the actual leg contact probability is less than the second preset probability value, the outer loop integral controller is disabled in the outer loop cycle, and the cumulative error of the outer loop integral controller is cleared to zero, so that the integral control output is zero. When the desired single-leg gait phase occurs at the moment of landing and the actual leg contact probability shows an increasing trend, the outer loop integral controller is disabled in that outer loop cycle, and the accumulated error of the outer loop integral controller is cleared to zero, so that the integral control output is zero.

[0011] In an optional implementation, each leg joint is in the first... Theoretical damping torque for one outer ring cycle The following formula is used to calculate: ; in, This is used to indicate that the joint of the leg entity is in the... Effective damping coefficient for each outer ring cycle This is used to indicate that the joint of the leg entity is in the... The actual joint angular velocity of one outer ring cycle This is used to indicate that the joint of the leg entity is in the... The expected joint angular velocity boundary corresponding to each outer loop cycle Used to represent proportional values The relevant damping scaling function.

[0012] In an optional implementation, for each traversed inner loop cycle, based on the target joint torque of the leg joint in the corresponding outer loop cycle, torque servo PI control is performed based on the desired feedforward torque, the desired single-leg gait phase, and the actual joint torque of the leg joint in that inner loop cycle to obtain the desired motor current of the leg joint in that inner loop cycle, including: The maximum motor current amplitude value of the leg joint in the inner loop cycle is determined based on the desired single-leg gait phase, and a corresponding current value constraint interval is constructed based on the maximum motor current amplitude value. Calculate the desired feedforward current corresponding to the desired feedforward torque, and calculate the joint torque error between the target joint torque and the actual joint torque corresponding to the inner loop period; Based on the joint torque error, current inner loop PI control is performed to obtain the reference motor current of the leg joint in the inner loop cycle. The reference motor current and the desired feedforward current are added together to obtain the corresponding candidate motor current. The candidate motor current is then truncated to the current value constraint range to obtain the desired motor current of the leg joint in the inner loop cycle.

[0013] In an optional implementation, the step of determining the maximum motor current amplitude value of the leg joint in the inner loop cycle based on the desired single-leg gait phase includes: The maximum current correction timer for the leg joint is detected to be active during the inner loop cycle. If the maximum current correction timer is detected to be active, the first preset current amplitude value is used as the maximum motor current amplitude value; otherwise, it is detected whether the desired single-leg gait phase belongs to the moment of landing. When the desired single-leg gait phase is detected to be at the moment of landing, the maximum current correction timer is activated and the first preset current amplitude value is used as the maximum motor current amplitude value. Otherwise, it is detected whether the desired single-leg gait phase is a support phase, whether the joint temperature rise rate of the leg joint exceeds a preset speed threshold, and whether the actual joint torque corresponding to the inner loop cycle is under high load. If the desired single-leg gait phase is detected to be a support phase, and the joint temperature rise rate exceeds the preset speed threshold, and the actual joint torque corresponding to the inner loop cycle is under high load, then the second preset current amplitude value is used as the maximum motor current amplitude value; otherwise, the rated current amplitude value of the motor of the leg joint is used as the maximum motor current amplitude value; wherein, the first preset current amplitude value is greater than the rated current amplitude value of the motor, and the rated current amplitude value of the motor is greater than the second preset current amplitude value.

[0014] Secondly, this application provides a multi-legged robot drive system, the system including a central controller and a plurality of joint controllers that are communicatively connected to each other, wherein each joint controller is used to control one leg joint of the multi-legged robot individually; The central controller works in conjunction with multiple joint controllers to implement the robot gait drive control method described in any of the foregoing embodiments for the multi-legged robot.

[0015] In this case, the beneficial effects of the embodiments of this application may include the following: This application utilizes a central controller to acquire the actual joint state data uploaded by multiple joint controllers (each joint controller is used to control one leg joint of the multi-legged robot) in the current control cycle, and acquires the torso IMU data, visual odometry data, and terrain perception data of the multi-legged robot in the current control cycle. Then, based on the acquired actual joint state data, torso IMU data, and visual odometry data, it determines the global motion state data of the multi-legged robot for the desired gait trajectory in the current control cycle. An adaptive control engine is then invoked to solve for leg joint motion commands based on the terrain perception data and global motion state data, according to the desired gait trajectory, to obtain the desired joint motion commands for each leg joint in the next control cycle, adapted to the terrain perception data. Next, the central controller sends the desired joint motion commands for each of the leg joints to the corresponding joint controllers. In the next control cycle, each joint controller performs force-position hybrid dual-loop PI control on the corresponding leg joint according to the received desired joint motion commands to drive the leg joint to move. This effectively improves the reliability of the robot drive system through the organic combination of distributed control architecture, multimodal perception mechanism, adaptive control algorithm, and force-position hybrid dual-loop control mechanism. It enables the robot to dynamically adjust its gait according to terrain changes in unstructured complex environments and improves the joint dynamic response accuracy during robot gait movement. This facilitates the achievement of smooth and adaptive gait movement of multi-legged robots in unstructured complex environments.

[0016] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the system composition of the multi-legged robot drive system provided in the embodiments of this application; Figure 2 A flowchart illustrating the robot gait drive control method provided in an embodiment of this application; Figure 3 for Figure 2 A schematic diagram illustrating the execution of step S230; Figure 4 for Figure 2 A schematic diagram illustrating the execution of step S240; Figure 5 for Figure 4 A schematic diagram illustrating the execution of neutron step S241; Figure 6 for Figure 4 A schematic diagram illustrating the execution of the neutron step S242.

[0019] Icons: 10 - Multi-legged robot drive system; 11 - Central controller; 12 - Joint controller. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0021] Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0022] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0023] In the description of this application, it should be understood that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product is in use, or the orientation or positional relationship commonly understood by those skilled in the art. They are used only for the convenience of describing this application and simplifying the description, and are not intended to indicate or imply that the equipment or component referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0024] In the description of this application, it should also be noted that, unless otherwise expressly specified and limited, the terms "set up," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0025] Furthermore, it is understood in the description of this application that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. Those skilled in the art will understand the specific meaning of the above terms in this application based on the specific circumstances.

[0026] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0027] Please refer to Figure 1 , Figure 1 This is a schematic diagram of the system composition of the multi-legged robot drive system 10 provided in this application embodiment. In this application embodiment, the multi-legged robot drive system 10 is deployed on the multi-legged robot to be driven (e.g., a Tyrannosaurus Rex bionic robot or a Brachiosaurus bionic robot) to drive the multi-legged robot to move. The multi-legged robot drive system 10 may include a central controller 11 and multiple joint controllers 12 to utilize a distributed control architecture to effectively separate high-level intelligent decision-making from high-real-time joint servo control, reducing the risk of single-point failure in the multi-legged robot drive system and improving the reliability of the robot drive system.

[0028] In this embodiment, the central controller 11 is responsible for implementing the gait behavior decision function of the multi-legged robot to be driven. The central controller 11 can be implemented using a high-performance embedded computing platform (e.g., NVIDIA Jetson series).

[0029] In this embodiment, each joint controller 12 is used to control the movement of one leg joint of the multi-legged robot to be driven. The multiple mechanical leg structures of the multi-legged robot to be driven each involve multiple leg joints. The central controller 11 is communicatively connected to all joint controllers 12 to send joint movement commands generated for different leg joints and matched with gait behavior decision results to the corresponding joint controllers 12. Each joint controller 12 performs high real-time joint servo control according to the received joint movement commands. Thus, through the organic cooperation between the central controller 11 and all joint controllers 12, the multi-legged robot to be driven can dynamically adjust its gait to follow terrain changes in unstructured complex environments, and improve the joint dynamic response accuracy during robot gait movement, making it easier to achieve a smooth and adaptive robot gait movement effect in unstructured complex environments. Each joint controller 12 can be implemented using a microcontroller unit (e.g., an STM32F4 / F7 series MCU). The central controller 11 is interconnected with each joint controller 12 via a CAN bus, which ensures real-time communication and anti-interference capabilities.

[0030] Understandable, Figure 1 The block diagram shown is only a schematic diagram of one component of the multi-legged robot drive system 10. The multi-legged robot drive system 10 may also include components such as... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown. Figure 1 The components shown can be implemented using hardware, software, or a combination thereof.

[0031] In this application, to ensure that the multi-legged robot drive system 10 can guarantee smooth and fluid adaptive gait movement of the multi-legged robot in complex unstructured environments, this application provides a robot gait drive control method to achieve the aforementioned objective. The robot gait drive control method provided in this application will be described in detail below.

[0032] Please refer to Figure 2 , Figure 2 This is a flowchart illustrating the robot gait drive control method provided in an embodiment of this application. In this embodiment, Figure 2 The robot gait drive control method shown is applied to the multi-legged robot drive system 10 described above. The robot gait drive control method may include steps S210 to S240.

[0033] In step S210, the central controller acquires the actual joint state data uploaded by each of the multiple joint controllers in the current control cycle, and acquires the torso IMU data, visual odometry data and terrain perception data of the multi-legged robot in the current control cycle.

[0034] In this embodiment, each joint controller 12 will upload the actual joint state data of the leg physical joint it is responsible for to the central controller 11 at the end of any control cycle (which records the actual joint angle, actual joint angular velocity and actual joint torque of the corresponding leg physical joint at the end of the control cycle, as well as the foot contact status of the mechanical leg structure to which the leg physical joint belongs).

[0035] In this embodiment, the multi-legged robot is equipped with various physical sensors, such as an IMU (integrated unit) deployed at the robot's torso (used to sense the robot's torso posture, torso angular velocity, and torso linear acceleration), an environmental perception sensor deployed at the robot's head or chest (which can be implemented using a depth camera or LiDAR to perform terrain perception of the actual working environment in which the multi-legged robot is located), and a visual odometer deployed at the robot's torso (used to measure the change in robot pose between any two adjacent control cycles). These physical sensors on the multi-legged robot also individually upload their sensed sensor data to the central controller 11 at the end of any control cycle. Examples include torso IMU data uploaded by the IMU (which records the actual robot torso posture, actual torso angular velocity, and actual torso linear acceleration at the end of the corresponding control cycle), visual odometer data uploaded by the visual odometer (which records the change in robot pose of the multi-legged robot in the corresponding control cycle relative to the previous control cycle), and terrain perception data uploaded by the environmental perception sensor.

[0036] In step S220, the central controller determines the global motion state data of the multi-legged robot for the desired gait trajectory in the current control cycle based on the acquired actual joint state data, torso IMU data and visual odometry data.

[0037] In this embodiment, the desired gait trajectory is a robot gait trajectory pre-configured by the robot operator for the multi-legged robot; the central controller 11 can use an extended Kalman filter algorithm to calculate the robot state data of the multi-legged robot at the end of the current control cycle in real time based on the actual joint state data of each leg joint, the acquired trunk IMU data, and visual odometry data. (This includes the global robot pose, three-dimensional velocity and three-dimensional acceleration of the center of mass of the multi-legged robot in the world coordinate system, the foot end pose of each mechanical leg structure in the robot base coordinate system, and the actual single-leg gait phase of each mechanical leg structure (which describes the corresponding mechanical leg structure in the robot gait phase). The system includes the gait phase state during the motion process (which can be the support phase, swing phase, or landing moment), and information such as the joint mapping pose, joint movement speed, and joint mapping torque of each leg joint in the robot's base coordinate system. Then, the central controller 11 determines the global tracking error data of the robot state data relative to the desired gait trajectory by comparing the robot state data with the desired gait trajectory, thereby obtaining the global motion state data of the multi-legged robot for the desired gait trajectory in the current control cycle (which consists of the robot state data of the multi-legged robot in the current control cycle and the global tracking error data of the multi-legged robot for the desired gait trajectory).

[0038] In step S230, the central controller calls the adaptive control engine to solve for the leg joint motion commands based on the terrain perception data and global motion state data and the desired gait trajectory, thereby obtaining the desired joint motion commands for each of the leg physical joints in the next control cycle.

[0039] In this embodiment, the adaptive control engine is built using a variety of advanced adaptive control algorithms, enabling it to include a gait planning and optimization network, a model predictive control module, and a sliding mode supervision module. The gait planning and optimization network is a lightweight policy network trained on a large scale using reinforcement learning algorithms, ensuring the adaptive control engine has the ability to learn from optimal gait experience data under different terrain conditions, allowing the multi-legged robot to automatically switch between optimal gait strategies in different terrains. The model predictive control module is built using the MPC (Model Predictive Control) algorithm, enabling the adaptive control engine to predict the future gait movement trends of each leg joint under the current terrain based on a variable stiffness single-rigid-body dynamics model. The sliding mode supervision module is built using a sliding mode algorithm, enabling the adaptive control engine to strongly suppress gait jitter caused by strong impacts, sudden terrain changes, and other disturbances, ensuring the multi-legged robot can quickly recover stability in unstructured environments. Therefore, this application can call the adaptive control engine to solve for the expected joint motion commands of each of the leg physical joints in the next control cycle that are adapted to the terrain perception data (which records the expected feedforward torque, expected joint angle, expected single-leg gait phase (which describes the expected gait phase state of the mechanical leg structure to which the corresponding leg physical joint belongs in the next control cycle), expected flutter suppression compensation torque, and expected joint angular velocity boundary (which describes the maximum joint angular velocity of the corresponding leg physical joint in the next control cycle) of the corresponding leg physical joint in the next control cycle).

[0040] Alternatively, please refer to Figure 3 , Figure 3 yes Figure 2 A schematic diagram illustrating the execution of step S230. In this embodiment of the application, step S230 may include sub-steps S231 to S234, which, through the deep fusion of multimodal perception mechanism and multimodal adaptive control algorithm, enable the robot to dynamically adjust its gait in unstructured environments to follow terrain changes during high-level intelligent decision-making.

[0041] Sub-step S231 calls the gait planning and optimization network to perform gait planning optimization inference based on robot state data, terrain perception data and desired gait trajectory, and obtains the desired gait parameter compensation amount, the model parameter optimization amount of the model prediction control module and the sliding mode gain fine-tuning amount of the sliding mode supervision module.

[0042] In this embodiment, the gait planning and optimization network can perform forward inference based on the robot state data, terrain perception data, and desired gait trajectory in a simulation environment to determine the model parameter optimization amount of the model prediction control module when adapting to the current terrain state (i.e., the model parameter correction amount of the strain stiffness single rigid body dynamics model (e.g., the correction amount for equivalent leg stiffness and dynamic damping coefficient)), the sliding mode supervision module's sliding mode gain fine-tuning amount when adapting to the current terrain state, and the desired gait parameter compensation amount of the multi-legged robot when adapting to the current terrain state (i.e., the parameter increment that needs to be compensated for in the current terrain state for robot gait parameters (e.g., leg lift height, stride length, step frequency, etc.)).

[0043] Sub-step S232: Call the model prediction control module to perform joint motion prediction for all leg joints based on the global motion state data according to the model parameter tuning amount, and obtain the theoretical feedforward torque, theoretical joint angle and theoretical single-leg gait phase of each leg joint in the next control cycle.

[0044] In this embodiment, the variable stiffness single rigid body dynamics model can be modified using the model parameter tuning amount. Then, the global motion state data is input into the modified variable stiffness single rigid body dynamics model. With the optimization objectives of minimizing energy consumption and maximizing stability margin, forward rolling time-domain optimization is performed based on the variable stiffness single rigid body dynamics model to predict the optimal joint motion state sequence of each of the leg joints involving multiple future control cycles (which records the theoretical feedforward torque, theoretical joint angle, and theoretical single-leg gait phase that the corresponding leg joint may achieve in each of the multiple future control cycles). The theoretical feedforward torque, theoretical joint angle, and theoretical single-leg gait phase of each of the leg joints in the first future control cycle (i.e., the next control cycle after the current control cycle) in the corresponding optimal joint motion state sequence are determined.

[0045] In sub-step S233, the sliding mode supervision module is invoked to perform chatter suppression prediction for all leg joints based on the global tracking error data according to the fine-tuning amount of the sliding mode gain, so as to obtain the theoretical chatter suppression compensation torque of each leg joint in the next control cycle.

[0046] In this embodiment, the sliding mode gain fine-tuning amount can be used to correct the sliding mode supervision module. Then, the sliding mode supervision module is invoked to calculate the sliding surface and its approach law that match the global tracking error data. Based on the calculated sliding surface and its approach law, it can be determined whether each leg joint touches the sliding surface under the current terrain condition. When it is determined that a certain leg joint touches the sliding surface, a non-zero theoretical chatter suppression compensation torque is adaptively generated for that leg joint to suppress the gait chattering problem currently generated by that leg joint. When a leg joint does not touch the sliding surface, the theoretical chatter suppression compensation torque for that leg joint in the next control cycle is zero.

[0047] Sub-step S234: Based on the expected gait parameter compensation amount and the expected gait trajectory, integrate the joint motion commands of each leg joint in the next control cycle with the theoretical feedforward torque, theoretical joint angle, theoretical single-leg gait phase and theoretical flutter suppression compensation torque, to obtain the expected joint motion commands of each leg joint in the next control cycle that satisfy the joint limit motion constraint conditions.

[0048] In this embodiment, the desired gait trajectory can be used as a macroscopic reference dominant vector. Using a joint motion solver based on time-optimal planning criteria and minimum acceleration criteria, under the premise of satisfying the joint limit motion constraints of each of the leg physical joints (which record the maximum joint angular velocity, maximum joint angular acceleration, and maximum joint torque of the corresponding leg physical joint at the physical hardware level), the theoretical feedforward torque, theoretical joint angle, theoretical single-leg gait phase, and theoretical flutter suppression compensation torque of each of the leg physical joints in the next control cycle are fused and optimized into multiple smooth and physically feasible joint motion commands, thereby obtaining the desired joint motion commands of each of the leg physical joints in the next control cycle.

[0049] Therefore, by executing the above sub-steps S231 to S234, this application can achieve the function of dynamic adjustment of robot gait following terrain changes in unstructured environments during high-level intelligent decision-making through the deep integration of multimodal perception mechanism and multimodal adaptive control algorithm.

[0050] In step S240, each joint controller obtains the desired joint motion command for the corresponding leg joint issued by the central controller, and performs force-position hybrid dual-loop PI control on the leg joint according to the desired joint motion command in the next control cycle to drive the leg joint to move.

[0051] In this embodiment, after the central controller 11 determines the expected joint motion command for each of the leg joints in the next control cycle, it sends the expected joint motion command corresponding to each leg joint to the corresponding joint controller 12. The joint controller 12 then implements a high real-time joint servo control effect for the leg joint based on the force-position hybrid dual closed-loop control mechanism according to the received expected joint motion command. Thus, through the organic cooperation between the central controller 11 and all joint controllers 12, a smooth and adaptive robot gait motion effect is achieved in unstructured complex environments.

[0052] Alternatively, please refer to Figure 4 , Figure 4 yes Figure 2 A schematic diagram of step S240 is shown. In this embodiment of the application, the step "each joint controller performs force-position hybrid dual-loop PI control on the corresponding leg joint according to the corresponding desired joint motion command in the next control cycle" in step S240 may include sub-steps S241 to S243, so that each joint controller 12 can use the force-position hybrid dual-loop control mechanism combined with local joint real-time sensing data (which may include the real-time joint angle, real-time joint angular velocity, real-time leg contact probability (which describes the probability of the foot end of the mechanical leg structure to which the corresponding leg joint belongs to contact the ground), real-time foot end three-dimensional force (i.e., the foot end three-dimensional force of the mechanical leg structure to which the corresponding leg joint belongs), real-time joint torque, etc.) to achieve a high real-time local intelligent servo control effect for the leg joint, ensuring the joint dynamic response accuracy and joint motion stability.

[0053] Sub-step S241: Iterate through the multiple outer loop cycles covered by the next control cycle in sequence, and for each outer loop cycle, perform force-position coordination PI control based on the actual joint angle, actual joint angular velocity, actual leg ground contact probability and actual foot three-dimensional force of the corresponding leg entity joint in the outer loop cycle, based on the expected joint angle, expected single-leg gait phase, expected flutter suppression compensation torque and expected joint angular velocity boundary, to obtain the target joint torque of the leg entity joint in the outer loop cycle.

[0054] In this embodiment, the force-position hybrid dual-closed-loop control mechanism, when applied to any control cycle, can divide the control cycle into multiple outer loop cycles, and further divide the multiple outer loop cycles under the same control cycle into multiple inner loop cycles. This allows each control cycle to independently cover multiple outer loop cycles, and ensures that each outer loop cycle can independently cover multiple inner loop cycles (for example, a 50ms control cycle can be divided into 50 1ms outer loop cycles, and each 1ms outer loop cycle can be converted into 5 20μs inner loop cycles). Each outer loop cycle utilizes a force-position coordinated PI (Proportion-Integral) control mechanism to quickly solve for the joint torque compensation amount (i.e., the target joint torque). Multiple outer loop cycles belonging to the same control cycle share the desired joint angle, desired single-leg gait phase, desired flutter suppression compensation torque, and desired joint angular velocity boundary of that control cycle.

[0055] Alternatively, please refer to Figure 5 , Figure 5 yes Figure 4 A schematic diagram illustrating the execution of sub-step S241. In this embodiment, sub-step S241 may include sub-steps S241a to S241d, so that the force-position coordination PI control loop can adaptively solve for the joint torque compensation amount following the single-leg gait phase change.

[0056] Sub-step S241a: Calculate the joint angle admittance offset of the leg entity joint in the outer ring cycle that matches the desired single-leg gait based on the actual three-dimensional force at the foot end.

[0057] In this embodiment, the joint angle admittance offset of each leg joint in any outer ring cycle is calculated using an admittance control algorithm based on the corresponding actual three-dimensional force at the foot end. The admittance control algorithm has different control parameters under different single-leg gait phases to dynamically adjust joint compliance according to changes in the single-leg gait phase, thereby avoiding joint hard impact and improving leg ground contact compliance. Specifically, each leg joint in the [missing information - likely a specific phase or cycle]... Joint angle admittance offset per outer loop cycle It can be calculated using the following formula: ; in, This is used to indicate that the joint of the leg entity is in the... The actual three-dimensional force at the foot during each outer cycle. Used to represent the desired foot force of the leg joint in the desired single-leg gait phase. The virtual joint mass used to represent the physical joint of the leg. Used to represent the actual duration of an outer loop period The virtual joint damping parameter is used to represent the actual joint of the leg in the desired single-leg gait phase. This parameter represents the virtual joint stiffness of the leg entity joint in the desired single-leg gait phase.

[0058] During this process, when the desired single-leg gait phase is the swing phase, both the virtual joint damping parameter and the virtual joint stiffness parameter are assigned smaller, stable values ​​to improve the joint compliance of the corresponding leg joint. When the desired single-leg gait phase is the support phase, the virtual joint damping parameter is set to its maximum value, while the virtual joint stiffness parameter is set to decrease from a larger value to a stable value to improve the joint angle tracking stiffness of the corresponding leg joint. When the desired single-leg gait phase is the landing moment, the virtual joint damping parameter is set to a larger value, while the virtual joint stiffness parameter is set to its maximum value.

[0059] Sub-step S241b calculates the joint angle error between the target joint angle and the actual joint angle in the outer loop cycle, and performs force-position outer loop PI control based on the joint angle error according to the expected single-leg gait phase and the actual leg ground contact probability, to obtain the reference joint torque of the leg entity joint in the outer loop cycle.

[0060] In this embodiment, the target joint angle of each leg joint in any outer loop cycle is obtained by adding the joint angle admittance offset of the leg joint in the corresponding outer loop cycle to the desired joint angle of the leg joint in the next control cycle. Each leg joint in the... The outer loop PI control law for each outer loop cycle is calculated using the following formula: ; in, This is used to indicate that the joint of the leg entity is in the... Reference joint torque for each outer ring cycle, This is used to indicate that the joint of the leg entity is in the... The joint angle error between the target joint angle and the actual joint angle in each outer ring cycle. Used to represent the proportional gain of the outer loop of the force potential. The outer-loop integral controller used to represent the leg joint is in the... Each outer ring cycle addresses joint angle errors. The proportional control output quantity; Used to represent the outer loop integral gain of the force potential. The outer-loop integral controller used to represent the leg joint is in the... Error of joint angle in the outer ring cycle The cumulative error, The outer-loop integral controller used to represent the leg joint is in the... Each outer ring cycle addresses joint angle errors. Integral control output quantity; The actual numerical value of the leg joint in the first... The expected single-leg gait phase of an outer loop cycle is related to the actual probability of leg contact with the ground.

[0061] Therefore, the above sub-step S241b may include: adjusting the start / stop state of the outer loop integral controller of the leg joint in the outer loop cycle according to the desired single-leg gait phase and the actual leg ground contact probability, and determining the integral control output of the outer loop integral controller for the joint angle error in the outer loop cycle; calculating the proportional control output of the outer loop proportional controller of the leg joint for the joint angle error in the outer loop cycle, and performing an addition operation on the proportional control output and integral control output corresponding to the outer loop cycle to obtain the reference joint torque of the leg joint in the outer loop cycle.

[0062] During this process, when the desired single-leg gait phase belongs to the support phase and the actual leg contact probability is greater than the first preset probability value (e.g., 0.8), the outer loop integral controller is activated in the outer loop cycle, and the outer loop integral controller is driven to generate the integral control output according to the nominal value of the outer loop integral gain, so that the foot part of the corresponding mechanical leg structure can stably land and achieve high-precision joint position tracking effect; when the desired single-leg gait phase belongs to the support phase and the actual leg contact probability is between the first preset probability value and the second preset probability value (wherein the second preset probability value is less than the first preset probability value, and the second preset probability value can be represented by 0.3), the outer loop integral controller is activated in the outer loop cycle, and the outer loop integral controller is driven to generate the integral control output according to the outer loop integral gain transition value (wherein the outer loop integral gain transition value is half of the nominal value of the outer loop integral gain), so that the corresponding mechanical leg structure avoids excessive accumulation of joint angle error in the early stage of foot contact.

[0063] When the desired single-leg gait phase is a swing phase and the actual leg contact probability is less than the second preset probability value, the outer-loop integral controller is disabled in that outer-loop cycle, and the accumulated error of the outer-loop integral controller is cleared to zero, making the integral control output zero, so that the corresponding mechanical leg structure is protected from the impact of the moment of contact. When the desired single-leg gait phase is a landing phase and the actual leg contact probability shows an increasing trend, the outer-loop integral controller is disabled in that outer-loop cycle, and the accumulated error of the outer-loop integral controller is cleared to zero, making the integral control output zero, so that the corresponding mechanical leg structure is protected from joint position deviation caused by the impact of contact. In other words, the outer-loop integral controller of each leg joint automatically clears the accumulated error each time it switches from the support phase to the swing phase.

[0064] Therefore, by executing the specific steps of the above sub-step S241b, this application can adaptively adjust the integral control output in accordance with the phase change of the single-leg gait during the force-position outer loop PI control process, so as to ensure that the solved reference joint torque can effectively match the phase change of the single-leg gait.

[0065] Sub-step S241c calculates the theoretical damping torque of the leg joint during the outer ring cycle based on the actual joint angular velocity and the expected joint angular velocity boundary.

[0066] In this embodiment, each leg joint is in the first... Theoretical damping torque for one outer ring cycle The following formula is used to calculate: ; in, This is used to indicate that the joint of the leg entity is in the... Effective damping coefficient for each outer ring cycle This is used to indicate that the joint of the leg entity is in the... The actual joint angular velocity of one outer ring cycle This is used to indicate that the joint of the leg entity is in the... The expected joint angular velocity boundary corresponding to each outer loop cycle Used to represent proportional values The relevant damping scaling function.

[0067] In one embodiment of this example, each leg joint is in the first... The effective damping coefficient of each outer loop cycle can be adaptively adjusted according to the single-leg gait phase change to ensure that the corresponding leg joint can effectively absorb the impact energy at the moment of foot landing. Based on this, the step of determining the effective damping coefficient of the leg joint in the traversed outer loop cycle according to the desired single-leg gait phase can include: detecting whether the damping coefficient correction timer of the leg joint is active in the outer loop cycle; if the damping coefficient correction timer is detected to be active, then a first preset damping coefficient is used as the effective damping coefficient; otherwise, detecting whether the desired single-leg gait phase belongs to the moment of landing; when the desired single-leg gait phase is detected to belong to the moment of landing, activating the damping coefficient correction timer and using the first preset damping coefficient as the effective damping coefficient; otherwise, using the nominal value of the damping coefficient as the effective damping coefficient. The damping coefficient correction timer, each time it is activated, will remain active for a subsequent period (e.g., 100ms); the first preset damping coefficient is greater than the nominal value of the damping coefficient.

[0068] Sub-step S241d involves adding the reference joint torque, theoretical damping torque, and desired flutter suppression compensation torque of the leg joint in the outer ring cycle to obtain the target joint torque of the leg joint in the outer ring cycle.

[0069] Therefore, this application can adaptively solve the joint torque compensation amount in the force-position coordination PI control link by executing the above sub-steps S241a to S241d, following the phase change of the single-leg gait.

[0070] Sub-step S242: Traverse the multiple inner loop cycles covered by the outer loop cycle, and for each traversed inner loop cycle, perform torque servo PI control based on the target joint torque of the leg joint in the outer loop cycle, the desired feedforward torque, the desired single-leg gait phase, and the actual joint torque of the leg joint in the corresponding inner loop cycle, to obtain the desired motor current of the leg joint in the inner loop cycle.

[0071] In this embodiment, multiple inner loop cycles belonging to the same outer loop cycle share the target joint torque of the outer loop cycle. The expected feedforward torque corresponding to the outer loop cycle bypasses the force-position coordination PI control loop and directly acts on the torque servo PI control loop of each inner loop cycle, so as to avoid increasing the calculation delay of the force-position coordination PI control loop and improve the joint dynamic response accuracy.

[0072] Alternatively, please refer to Figure 6 , Figure 6 yes Figure 4A schematic diagram illustrating the execution of sub-step S242. In this embodiment, sub-step S242 may include sub-steps S242a to S242d, so that the torque servo PI control loop can adaptively adjust the desired joint motor current of the corresponding leg joint in any inner loop cycle according to the phase change of the single-leg gait, and achieve the intelligent protection effect of the motor during shutdown.

[0073] Sub-step S242a: Determine the maximum motor current amplitude value of the leg entity joint in the inner loop cycle according to the desired single-leg gait phase, and construct the corresponding current value constraint interval based on the maximum motor current amplitude value.

[0074] In this embodiment, the lower limit of the current value constraint interval is the opposite of the maximum motor current amplitude value, and the upper limit of the current value constraint interval is the maximum motor current amplitude value. The step "determine the maximum motor current amplitude value of the leg joint in the inner loop cycle based on the desired single-leg gait phase" in sub-step S242a may include: The maximum current correction timer for the leg joint is detected to be active during the inner loop cycle. If the maximum current correction timer is detected to be active, the first preset current amplitude value is used as the maximum motor current amplitude value; otherwise, it is detected whether the desired single-leg gait phase belongs to the moment of landing. When the desired single-leg gait phase is detected to be at the moment of landing, the maximum current correction timer is activated and the first preset current amplitude value is used as the maximum motor current amplitude value. Otherwise, it is detected whether the desired single-leg gait phase is a support phase, whether the joint temperature rise rate of the leg joint exceeds a preset speed threshold, and whether the actual joint torque corresponding to the inner loop cycle is under high load. If the desired single-leg gait phase is detected to be a support phase, and the joint temperature rise rate exceeds the preset speed threshold, and the actual joint torque corresponding to the inner loop cycle is under high load, then the second preset current amplitude value is used as the maximum motor current amplitude value; otherwise, the rated current amplitude value of the motor of the leg joint is used as the maximum motor current amplitude value; wherein, the first preset current amplitude value is greater than the rated current amplitude value of the motor, and the rated current amplitude value of the motor is greater than the second preset current amplitude value.

[0075] During this process, the maximum current correction timer will remain active for a period of time (e.g., 100ms) each time it is activated, so that the maximum motor current amplitude of the corresponding leg joint can remain higher than the rated motor current amplitude during the timer activation period, so as to effectively offset the impact of the landing moment; the second preset current amplitude value is used as the maximum motor current amplitude value to prevent the corresponding leg joint from running under high load for a long time and to avoid the leg joint from overheating and shutting down.

[0076] Therefore, by executing the specific steps of the above sub-step S242a, this application enables the constructed current value constraint range to adaptively select appropriate shutdown motor intelligent protection operation in accordance with the phase change of single-leg gait.

[0077] Sub-step S242b calculates the desired feedforward current corresponding to the desired feedforward torque, and calculates the joint torque error between the target joint torque and the actual joint torque corresponding to the inner loop period.

[0078] In this embodiment, the desired feedforward current is obtained by dividing the desired feedforward torque by the motor torque constant of the corresponding leg joint.

[0079] Sub-step S242c involves performing inner-loop PI control based on the joint torque error to obtain the reference motor current of the leg joint in that inner-loop cycle.

[0080] In this embodiment, each leg joint is in the first... The first outer ring cycle covered by the The current inner-loop PI control law for each inner loop cycle is calculated using the following formula: ; in, This is used to indicate that the joint of the leg entity is in the... The first outer ring cycle covered by the Reference motor current for each inner loop cycle, This is used to indicate that the joint of the leg entity is in the... The first outer ring cycle covered by the Joint torque error of one inner loop cycle Used to represent the proportional gain of the inner current loop. The inner-loop integral controller used to represent the joint of the leg body is in the... The first outer ring cycle covered by the The output quantity is controlled proportionally within each inner loop cycle; Used to represent the inner loop integral gain of the current. Used to represent the duration of an inner loop period. The inner-loop integral controller used to represent the joint of the leg body is in the... The first outer ring cycle covered by the The integral control output of each inner loop cycle.

[0081] In sub-step S242d, the reference motor current and the desired feedforward current are added together to obtain the corresponding candidate motor current. The candidate motor current is then truncated to the current value constraint range to obtain the desired motor current of the leg joint in the inner loop cycle.

[0082] Therefore, by executing the above sub-steps S242a to S242d, this application can follow the phase change of the single-leg gait in the torque servo PI control link, adaptively adjust the expected joint motor current of the corresponding leg physical joint in any inner loop cycle, and achieve the intelligent protection effect of the motor when shutting down.

[0083] In sub-step S243, the joint motor driving the leg joint operates according to the corresponding desired motor current during the inner loop cycle.

[0084] In this embodiment, a joint controller determines that the corresponding leg joint is in the [missing information]. The first outer ring cycle covered by the After the expected motor current of the inner loop cycle, the leg joint will be driven in the first... Within each inner loop cycle, the motor operates according to the desired motor current, so that by introducing a force-position hybrid dual closed-loop control mechanism during the joint servo control process, the dynamic response accuracy of the joint and the real-time performance of the joint servo control can be effectively improved.

[0085] Therefore, by executing the above sub-steps S241 to S243, this application enables each joint controller 12 to utilize the force-position hybrid dual closed-loop control mechanism combined with local joint real-time sensing data to achieve a high real-time local intelligent servo control effect for the leg joint under its responsibility, ensuring the accuracy of joint dynamic response and the stability of joint movement.

[0086] Meanwhile, by executing the above steps S210 to S240, this application introduces a distributed control architecture to achieve effective separation between high-level intelligent decision-making and high-real-time joint servo control, reducing the risk of single-point failure in the multi-legged robot drive system 10, improving the reliability of the drive system, and realizing the gait dynamic adjustment function of following terrain changes in unstructured environments through the deep integration of multimodal perception mechanism and adaptive control algorithm in the high-level intelligent decision-making process. At the same time, by introducing a force-position hybrid dual closed-loop control mechanism in the joint servo control process, the dynamic response accuracy of the joint is effectively improved, thereby ensuring the smooth and adaptive gait movement of the multi-legged robot in complex unstructured environments.

[0087] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the architecture, functionality, and operation of possible implementations of the apparatus, methods, and computer program products according to embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0088] The above descriptions are merely various embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A robot gait drive control method, characterized in that, A method for a multi-legged robot drive system, the system comprising a central controller and multiple joint controllers interconnected, wherein each joint controller is individually used to control one leg joint of the multi-legged robot, the method comprising: The central controller acquires the actual joint state data uploaded by each of the multiple joint controllers in the current control cycle, and acquires the torso IMU data, visual odometry data and terrain perception data of the multi-legged robot in the current control cycle. The central controller determines the global motion state data of the multi-legged robot for the desired gait trajectory in the current control cycle based on the acquired actual joint state data, trunk IMU data, and visual odometry data. The central controller calls the adaptive control engine to solve for the leg joint motion commands based on the terrain perception data and the global motion state data, and the expected gait trajectory, so as to obtain the expected joint motion commands of each leg joint in the next control cycle. Each joint controller receives the desired joint motion command for the corresponding leg joint from the central controller, and performs force-position hybrid dual-loop PI control on the leg joint according to the desired joint motion command in the next control cycle to drive the leg joint to move.

2. The method according to claim 1, characterized in that, The global motion state data includes the robot state data of the multi-legged robot and the global tracking error data of the multi-legged robot for the desired gait trajectory. The adaptive control engine includes a gait planning and optimization network, a model prediction control module, and a sliding mode supervision module. The central controller calls the adaptive control engine to solve for the leg joint motion commands based on the terrain perception data and the global motion state data, and obtains the desired joint motion commands for each leg joint in the next control cycle. The steps include: The gait planning and optimization network is invoked to perform gait planning optimization inference based on the robot state data, the terrain perception data, and the desired gait trajectory, to obtain the desired gait parameter compensation amount, the model parameter optimization amount of the model prediction control module, and the sliding mode gain fine-tuning amount of the sliding mode supervision module; The model prediction control module is invoked to perform joint motion prediction for all leg joints based on the global motion state data according to the model parameter tuning amount, so as to obtain the theoretical feedforward torque, theoretical joint angle and theoretical single-leg gait phase of each leg joint in the next control cycle. The sliding mode supervision module is invoked to perform chatter suppression prediction for each of the leg joints based on the global tracking error data according to the sliding mode gain fine-tuning amount, so as to obtain the theoretical chatter suppression compensation torque of each of the leg joints in the next control cycle. Based on the expected gait parameter compensation amount and the expected gait trajectory, the theoretical feedforward torque, theoretical joint angle, theoretical single-leg gait phase and theoretical flutter suppression compensation torque of each of the leg joints in the next control cycle are integrated to obtain the expected joint movement commands of each of the leg joints in the next control cycle that satisfy the joint limit movement constraints.

3. The method according to claim 1 or 2, characterized in that, The desired joint motion command for each leg joint in the next control cycle includes the desired feedforward torque, desired joint angle, desired single-leg gait phase, desired flutter suppression compensation torque, and desired joint angular velocity boundary for the corresponding leg joint. Each control cycle covers multiple outer loop cycles, and each outer loop cycle covers multiple inner loop cycles. The steps for each joint controller to perform force-position hybrid dual-loop PI control on the corresponding leg joint according to the desired joint motion command in the next control cycle include: The system sequentially traverses multiple outer loop cycles covered by the next control cycle. For each traversed outer loop cycle, based on the actual joint angle, actual joint angular velocity, actual leg ground contact probability, and actual foot three-dimensional force of the corresponding leg joint in that outer loop cycle, force-position coordination PI control is performed based on the desired joint angle, the desired single-leg gait phase, the desired flutter suppression compensation torque, and the desired joint angular velocity boundary to obtain the target joint torque of the leg joint in that outer loop cycle. The outer loop cycle covers multiple inner loop cycles, and for each inner loop cycle, based on the target joint torque of the leg joint in the outer loop cycle, torque servo PI control is performed based on the expected feedforward torque, the expected single-leg gait phase, and the actual joint torque of the leg joint in the corresponding inner loop cycle to obtain the expected motor current of the leg joint in the inner loop cycle. The joint motor that drives the leg joint operates according to the corresponding desired motor current during the inner loop cycle.

4. The method according to claim 3, characterized in that, For each traversed outer loop cycle, the step of performing force-position coordination PI control based on the expected joint angle, expected single-leg gait phase, expected flutter suppression compensation torque, and expected joint angular velocity boundary of the corresponding leg entity joint in that outer loop cycle to obtain the target joint torque of the leg entity joint in that outer loop cycle includes: Calculate the joint angle admittance offset of the leg joint in the outer circumference period that matches the desired single-leg gait based on the actual three-dimensional force at the foot end. Calculate the joint angle error between the target joint angle and the actual joint angle in the outer loop cycle, and perform force-position outer loop PI control based on the joint angle error according to the expected single-leg gait phase and the actual leg ground contact probability to obtain the reference joint torque of the leg entity joint in the outer loop cycle. The target joint angle in the outer loop cycle is obtained by adding the joint angle admittance offset of the corresponding outer loop cycle and the expected joint angle. Based on the actual joint angular velocity and the desired joint angular velocity boundary, calculate the theoretical damping torque of the leg joint in the outer ring cycle; The reference joint torque, theoretical damping torque, and desired flutter suppression compensation torque of the leg joint in the outer ring cycle are added together to obtain the target joint torque of the leg joint in the outer ring cycle.

5. The method according to claim 4, characterized in that, The step of performing force-position outer loop PI control based on the expected single-leg gait phase and the actual leg ground contact probability, and obtaining the reference joint torque of the leg's physical joint in the outer loop cycle, includes: Based on the desired single-leg gait phase and the actual leg contact probability, the start and stop states of the outer loop integral controller of the leg joint in the outer loop cycle are adjusted, and the integral control output of the outer loop integral controller for the joint angle error in the outer loop cycle is determined. The proportional control output of the outer loop proportional controller of the leg joint is calculated for the joint angle error in the outer loop cycle, and the proportional control output and integral control output corresponding to the outer loop cycle are added to obtain the reference joint torque of the leg joint in the outer loop cycle.

6. The method according to claim 5, characterized in that, The step of adjusting the start / stop state of the outer loop integral controller of the leg joint in the outer loop cycle according to the desired single-leg gait phase and the actual leg ground contact probability, and determining the integral control output of the outer loop integral controller for the joint angle error in the outer loop cycle, includes: When the desired single-leg gait phase belongs to the support phase and the actual leg contact probability is greater than the first preset probability value, the outer loop integral controller is activated in the outer loop cycle, and the outer loop integral controller is driven to generate the integral control output according to the nominal value of the outer loop integral gain. When the desired single-leg gait phase belongs to the support phase and the actual leg contact probability is between the first preset probability value and the second preset probability value, the outer loop integral controller is activated in the outer loop cycle, and the outer loop integral controller is driven to generate the integral control output according to the outer loop integral gain transition value, wherein the second preset probability value is less than the first preset probability value, and the outer loop integral gain transition value is half of the nominal value of the outer loop integral gain. When the desired single-leg gait phase is a swing phase and the actual leg contact probability is less than the second preset probability value, the outer loop integral controller is disabled in the outer loop cycle, and the cumulative error of the outer loop integral controller is cleared to zero, so that the integral control output is zero. When the desired single-leg gait phase occurs at the moment of landing and the actual leg contact probability shows an increasing trend, the outer loop integral controller is disabled in that outer loop cycle, and the accumulated error of the outer loop integral controller is cleared to zero, so that the integral control output is zero.

7. The method according to claim 4, characterized in that, Each leg joint in the first Theoretical damping torque for one outer ring cycle The following formula is used to calculate: ； in, This is used to indicate that the joint of the leg entity is in the... Effective damping coefficient for each outer ring cycle This is used to indicate that the joint of the leg entity is in the... The actual joint angular velocity of one outer ring cycle This is used to indicate that the joint of the leg entity is in the... The expected joint angular velocity boundary corresponding to each outer loop cycle Used to represent ratio values The relevant damping scaling function.

8. The method according to claim 3, characterized in that, For each traversed inner loop cycle, based on the target joint torque of the leg joint in the corresponding outer loop cycle, torque servo PI control is performed based on the desired feedforward torque, the desired single-leg gait phase, and the actual joint torque of the leg joint in that inner loop cycle to obtain the desired motor current of the leg joint in that inner loop cycle. This includes the following steps: The maximum motor current amplitude value of the leg joint in the inner loop cycle is determined based on the desired single-leg gait phase, and a corresponding current value constraint interval is constructed based on the maximum motor current amplitude value. Calculate the desired feedforward current corresponding to the desired feedforward torque, and calculate the joint torque error between the target joint torque and the actual joint torque corresponding to the inner loop period; Based on the joint torque error, current inner loop PI control is performed to obtain the reference motor current of the leg joint in the inner loop cycle. The reference motor current and the desired feedforward current are added together to obtain the corresponding candidate motor current. The candidate motor current is then truncated to the current value constraint range to obtain the desired motor current of the leg joint in the inner loop cycle.

9. The method according to claim 8, characterized in that, The step of determining the maximum motor current amplitude value of the leg joint in the inner loop cycle based on the desired single-leg gait phase includes: The maximum current correction timer for the leg joint is detected to be active during the inner loop cycle. If the maximum current correction timer is detected to be active, the first preset current amplitude value is used as the maximum motor current amplitude value; otherwise, it is detected whether the desired single-leg gait phase belongs to the moment of landing. When the desired single-leg gait phase is detected to be at the moment of landing, the maximum current correction timer is activated and the first preset current amplitude value is used as the maximum motor current amplitude value. Otherwise, it is detected whether the desired single-leg gait phase is a support phase, whether the joint temperature rise rate of the leg joint exceeds a preset speed threshold, and whether the actual joint torque corresponding to the inner loop cycle is under high load. If the desired single-leg gait phase is detected to be a support phase, and the joint temperature rise rate exceeds the preset speed threshold, and the actual joint torque corresponding to the inner loop cycle is under high load, then the second preset current amplitude value is used as the maximum motor current amplitude value; otherwise, the rated current amplitude value of the motor of the leg joint is used as the maximum motor current amplitude value; wherein, the first preset current amplitude value is greater than the rated current amplitude value of the motor, and the rated current amplitude value of the motor is greater than the second preset current amplitude value.

10. A multi-legged robot drive system, characterized in that, The system includes a central controller and multiple joint controllers that are interconnected, wherein each joint controller is used to control one leg joint of the multi-legged robot. The central controller works in conjunction with multiple joint controllers to implement the robot gait drive control method according to any one of claims 1-9 for the multi-legged robot.

Images