Motor drive control system, electric drive joint and control method thereof

Abstract

The application discloses a motor driving control system, an electric driving joint and a control method thereof. The control system comprises a first control module and a second control unit. The first control module comprises a first control unit, a signal gating control unit and a motor unit connected in sequence. The first control unit is configured to generate a first control signal according to the operation feedback signal of the motor unit through a vector control algorithm. The second control unit is electrically connected with the signal gating control unit. The second control unit is configured to generate a second control signal according to the operation feedback signal of the motor unit through a reinforcement learning algorithm. The signal gating control unit is configured to select the first control signal or the second control signal as a target control signal to control the operation of the motor unit according to the operation feedback signal of the motor unit and the second control signal. The application can solve the problem that the electric driving joint cannot quickly and safely and reliably adapt to different humanoid robots and has poor adaptability.

Description

Technical Field

[0001] This application relates to the field of motor control technology, and more specifically, to a motor drive control system, an electric drive joint, and a control method thereof. Background Technology

[0002] In related technologies, electrically driven joints are mainly used in biomimetic devices such as humanoid robots, robot dogs, and robotic arms. The motors in the electrically driven joints of humanoid robots are primarily controlled based on traditional control circuits and software architectures, such as traditional vector control algorithms. To improve control performance, a precise physical and mathematical model corresponding to the electrically driven joint and the robot's load needs to be established during control. The parameters in this physical and mathematical model are then used to adjust the joint motor control parameters. However, due to differences in their structure, degrees of freedom, load characteristics, and motion conditions, different humanoid robots exhibit significant differences in the electrical characteristics, mechanical characteristics, and inertia parameters of their corresponding electrically driven joints. This means that when using vector control algorithms to control the motors, the electrically driven joints applied to other humanoid robots and operating under different conditions require re-establishing or calibrating the model and debugging, which takes a considerable amount of time. Furthermore, during operation, the electrically driven joints lack the ability to self-optimize performance and efficiency based on real-time or short-term data, resulting in poor adaptive and self-matching performance.

[0003] Joint motor control based on reinforcement learning algorithms learns optimal control strategies autonomously through interaction with the environment, without relying on precise mathematical models. This results in strong adaptability and robustness, addressing the issue of strong dependence on joint and load parameters inherent in vector control algorithms. However, applying reinforcement learning to humanoid robot joint motor control requires self-learning training based on a period of operational data, a process currently typically conducted in simulations. Due to the discrepancy between simulation and reality, even after simulation adaptation, joint motor control applied to humanoid robots still suffers from control instability, necessitating iterative corrections.

[0004] Therefore, how to make the electric drive joint motor control system have strong self-adaptability (intelligent and with strong generalization) while adapting to and combining with scenario applications to ensure the safe and reliable operation of the humanoid robot is a key issue that industry representatives are focusing on and solving. Summary of the Invention

[0005] The main objective of this application is to provide a motor drive control system, an electric drive joint, and a control method thereof to solve the problems of electric drive joints being unable to quickly, safely, and reliably achieve fully autonomous adaptation and poor adaptability when applied to different humanoid robots, and to achieve the effect of meeting the requirements of high safety in humanoid robot application scenarios during debugging and operation.

[0006] According to one aspect of this application, a motor drive control system is provided, comprising: The first control module includes a first control unit, a signal gating control unit, and a motor unit that are electrically connected in sequence. The first control unit is configured to generate a first control signal based on the operation feedback signal of the motor unit through a vector control algorithm. The second control unit is electrically connected to the signal gating control unit. The second control unit is configured to operate in parallel with the first control unit and generate a second control signal based on the operation feedback signal of the motor unit through a reinforcement learning algorithm. The signal selection control unit is configured to select either the first control signal or the second control signal as the target control signal based on the operation feedback signal of the motor unit and the second control signal, and control the operation of the motor unit based on the target control signal.

[0007] Furthermore, the signal gating control unit is configured such that: when the motor unit starts running, the signal gating control unit selects the first control signal as the target control signal; when the motor unit reaches a stable operating state and the second control signal converges, the signal gating control unit selects the second control signal as the target control signal.

[0008] Furthermore, the motor drive control system also includes a torque shutdown module, which is electrically connected to the first control unit and the signal gating control unit respectively. The torque shutdown module is configured to cut off the output torque of the motor unit according to the torque shutdown signal issued by the first control unit and / or the signal gating control unit.

[0009] Furthermore, the motor drive control system also includes a brake locking module, which includes a brake locking logic circuit and a locking mechanism. The locking mechanism is disposed on the motor unit. The brake locking logic circuit is electrically connected to the first control unit, the signal gating control unit, and the locking mechanism, respectively. The brake locking logic circuit is configured to control the locking mechanism to lock the motor unit according to the brake locking signal issued by the first control unit and / or the signal gating control unit.

[0010] Furthermore, the first control unit includes at least one of an ARM chip, an MCU chip, and a DSP chip; and / or, The second control unit includes an AI computing chip; and / or, The signal gating control unit includes an FPGA chip.

[0011] According to another aspect of this application, an electrically driven joint structure is also provided, the electrically driven joint structure including the above-described motor drive control system and joint module, wherein the motor unit in the motor drive control system and the joint module are drivenly connected.

[0012] According to another aspect of this application, a control method for an electrically driven joint is also provided, applicable to the above-described electrically driven joint structure, the control method comprising: Step S1: After the electric drive joint is installed on the predetermined device, the debugging mode is started. The first control unit is controlled to generate a first control signal based on the vector control algorithm and the operation feedback signal of the motor unit. The second control unit is controlled to run in parallel with the first control unit and generate a second control signal based on the reinforcement learning algorithm and the operation feedback signal of the motor unit. Step S2: The signal gating control unit receives the first control signal, the second control signal, and the operation feedback signal of the motor unit, controls the operation of the motor unit through the first control signal, and at the same time determines whether the second control signal converges; Step S3: If the second control signal does not converge or the electric drive joint does not reach a stable operating state, the motor unit continues to be controlled by the first control signal; if the second control signal converges and the electric drive joint reaches a stable operating state, the signal gating control unit controls the motor unit to operate by the second control signal. Step S4: After the electric drive joint reaches a stable operating state under the control of the second control signal, the debugging mode is turned off. Based on the first control signal and the second control signal, the motor unit is controlled to operate in a preset operating mode to make the electric drive joint move.

[0013] Furthermore, the preset operating mode includes: A first operating mode is configured to use the second control signal as the target control signal and use the first control signal as a monitoring comparison signal to monitor and adjust the second control signal in real time. The second operating mode is configured to use the first control signal as the target control signal, and simultaneously calculate the control parameter correction amount corresponding to the load change or inertia change of the electric drive joint through the second control signal, and optimize the first control signal based on the control parameter correction amount.

[0014] Furthermore, the control method further includes: When torque shutdown of the electrically driven joint is required, the torque shutdown module receives a torque shutdown signal synchronously issued by the first control unit and the signal gating control unit, and cuts off the output torque between the motor unit and the joint module based on the torque shutdown signal; and / or, When it is necessary to brake and lock the electrically driven joint, the brake locking module receives the brake locking signal synchronously sent by the first control unit and the signal gating control unit, and drives the locking mechanism to brake and lock the electrically driven joint based on the brake locking signal.

[0015] Furthermore, the operating feedback signal of the motor unit includes at least one of the following: motor phase current, DC bus voltage, motor speed, and motor position, wherein: The step of generating the first control signal based on the vector control algorithm combined with the operating feedback signal of the motor unit includes: constructing a motor closed-loop control model using the motor phase current, the DC bus voltage, the motor speed, and the motor position as control measurement signals; and generating the first control signal through the motor closed-loop control model; and / or, The step of generating the second control signal based on the reinforcement learning algorithm and the operating feedback signal of the motor unit includes: using the motor speed, the motor phase current, and the DC bus voltage as state parameters, and the motor control command as action parameters, constructing a reinforcement learning model by maximizing a preset reward function to learn the optimal control strategy, and outputting the second control signal through the reinforcement learning model.

[0016] This application employs a dual-control architecture combining vector control and reinforcement learning control. The reinforcement learning algorithm of the second control unit can autonomously iteratively optimize the control strategy based on motor operation feedback signals, eliminating the need for rebuilding mathematical models and debugging. This significantly reduces the adaptation cost of porting electric drive joints across different robot models and improves the system's versatility. The vector control algorithm of the first control unit has a mature theoretical foundation and can provide stable and accurate basic control outputs during the steady-state operation of the motor, ensuring the basic motion performance of the electric drive joint. The reinforcement learning algorithm of the second control unit can better handle complex nonlinear and strongly coupled working conditions. In scenarios with sudden changes in joint load and complex and variable motion trajectories, it can generate better control signals through real-time feedback iteration, compensating for the limitations of vector control in dynamic conditions. The signal gating control unit intelligently switches control strategies based on motor operation feedback signals and the second control signal, achieving a combination of steady-state precise control and dynamic adaptive optimization, improving the control accuracy and response speed of the electric drive joint under all working conditions. Furthermore, the parallel operation architecture of the two control units has redundancy characteristics. When one control signal has insufficient adaptability or malfunctions, the signal gating control unit can quickly switch to another control signal, preventing the electric drive joint from going out of control. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, are illustrative and descriptive, serving to explain this application and do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the structure of a motor drive control system disclosed in an embodiment of this application; Figure 2 This is a detailed structural schematic diagram of a motor drive control system disclosed in an embodiment of this application; Figure 3 This is a flowchart illustrating a control method for an electrically driven joint disclosed in an embodiment of this application; Figure 4 This is a schematic diagram illustrating the principle of controlling the electrically driven joint using a vector control algorithm as disclosed in an embodiment of this application. Figure 5 This is a schematic diagram illustrating the principle of controlling an electrically driven joint using a reinforcement learning algorithm, as disclosed in an embodiment of this application.

[0018] The above figures include the following reference numerals: 10. Second control unit; 11. AI computing chip; 20. First control unit; 30. Signal gating control unit; 40. Motor unit; 50. Data acquisition module; 51. AD high-speed acquisition circuit; 60. Torque shutdown module; 61. Torque shutdown logic circuit; 70. Brake locking module; 71. Brake locking logic circuit; 72. Locking mechanism; 80. Drive circuit; 90. Power device. Detailed Implementation

[0019] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0020] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0021] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the invention. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale. Techniques, methods, and devices known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and devices should be considered part of the specification. In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters in the following figures denote similar items; therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0022] As described in the background section, in related technologies, the motors in the electric drive joints of humanoid robots are mainly controlled based on traditional vector control algorithms. During control, it is necessary to establish an accurate physical mechanism mathematical model corresponding to the electric drive joint. However, due to differences in their own structure, degrees of freedom, load characteristics, and motion conditions, different humanoid robots require the electric drive joint to be re-established or calibrated and debugged when the motor is controlled based on vector control algorithms for other humanoid robots and working under different conditions. Debugging requires a lot of time, and during operation, the electric drive joint lacks the function of self-optimizing performance and efficiency based on real-time or a period of working condition data, resulting in poor adaptive and self-matching performance.

[0023] Joint motor control based on reinforcement learning algorithms learns optimal control strategies autonomously through interaction with the environment, without relying on precise mathematical models. This results in strong adaptability and robustness, addressing the issue of strong dependence on joint and robot load parameters inherent in vector control algorithms. However, applying reinforcement learning to humanoid robot joint motor control requires self-learning training based on a period of operational data, a process currently typically conducted in simulations. Due to the discrepancy between simulation and reality, even after simulation adaptation, joint motor control applied to humanoid robots still suffers from control instability, necessitating iterative corrections.

[0024] Therefore, how to make the electric drive joint motor control system have strong self-adaptability (intelligent and with strong generalization) while adapting (the debugging process) and combining it with the application scenario to ensure the safe and reliable operation of the humanoid robot is a key issue that industry representatives are focusing on and solving.

[0025] To address this, this application provides a motor drive control system, an electric drive joint, and a control method thereof. This system, along with the electric drive joint and its control method, utilizes a reinforcement learning-based second control unit. The signal gating control unit selects a target control signal to control the motor unit's operation based on the motor unit's operational feedback signal and a second control signal issued by the second control unit. This solves the problems of electric drive joints failing to achieve rapid, safe, and reliable fully autonomous adaptation and exhibiting poor adaptability when applied to different humanoid robots. Furthermore, it achieves the effect of meeting the high-safety requirements of humanoid robot application scenarios during both debugging and operation. The following description, in conjunction with the accompanying drawings, will illustrate the motor drive control system, electric drive joint, and its control method of this application.

[0026] See Figure 1 and Figure 2 As shown in the figure, this application provides a motor drive control system, including a first control module and a second control unit 10. The first control module includes a first control unit 20, a signal gating control unit 30, and a motor unit 40, which are electrically connected in sequence. The first control unit 20 is configured to generate a first control signal based on the operating feedback signal of the motor unit 40 using a vector control algorithm. The second control unit 10 is electrically connected to the signal gating control unit 30 and is configured to operate in parallel with the first control unit 20, and generate a second control signal based on the operating feedback signal of the motor unit 40 using a reinforcement learning algorithm. The signal gating control unit 30 is configured to select either the first control signal or the second control signal as the target control signal based on the operating feedback signal of the motor unit 40 and the second control signal, and control the operation of the motor unit 40 based on the target control signal.

[0027] It is understood that the motor drive control system in this embodiment is mainly applied to the electric drive joints of the humanoid robot, that is, the motor drive system can drive the movement of the electric drive joints of the humanoid robot, thereby enabling the humanoid robot to move.

[0028] Specifically, in this embodiment, when the motor drive control system is applied to an electrically driven joint, the first control unit 20 processes the operation feedback signal of the motor unit 40 using a vector control algorithm, generates a first control signal, and sends it to the signal selection control unit 30. Simultaneously with the operation of the first control unit 20, the second control unit 10 operates in parallel with it. The second control unit 10 processes the operation feedback signal of the motor unit 40 using a reinforcement learning algorithm, generates a second control signal, and sends it to the signal selection control unit 30. Upon receiving the first and second control signals, the signal selection control unit judges the operation feedback signal of the motor unit 40 and the second control signal, selecting one as the target control signal for controlling the operation of the motor unit 40. The target control signal can be set as, for example, a PWM duty cycle, a voltage vector, or a modulation signal.

[0029] In this embodiment, the multi-channel parallel control mode of the first control unit 20 and the second control unit 10 can realize intelligent control of the motor and improve the safety of the electric drive joint motion control. The second control unit 10 generates the second control signal through reinforcement learning algorithm, which makes the control system have strong generalization ability. This allows the electric drive joint using the motor drive control system to be better adapted when applied to different humanoid robots. It can autonomously adapt to the operating conditions of the humanoid robot without the need for manual debugging, and can meet the requirements of high-safety humanoid robot application scenarios during operation.

[0030] Furthermore, the motor drive control system of this application embodiment also includes a data acquisition module 50. The data acquisition module 50 is electrically connected to the first control unit 20, the second control unit 10, and the signal gating control unit 30, respectively. The data acquisition module 50 is used to acquire the real-time operation feedback signal of the motor unit 40 and send it to the first control unit 20, the second control unit 10, and the signal gating control unit 30, respectively, so as to ensure that the control of the first control unit 20, the second control unit 10, and the signal gating control unit 30 can be based on the operation feedback signal of the motor unit 40, that is, the real-time operation data of the motor unit 40.

[0031] like Figure 2As shown, the data acquisition module 50 in this embodiment includes an AD high-speed acquisition circuit 51. The high-speed acquisition characteristics of the AD high-speed acquisition circuit 51 can accurately capture the dynamic parameter details such as current, speed, and position during the real-time operation of the motor, avoiding the situation where the control signal is out of sync with the actual operating condition of the motor due to signal acquisition lag or distortion. At the same time, the high-precision data acquisition capability can provide reliable raw data support for the vector control algorithm of the first control unit 20 and the reinforcement learning algorithm of the second control unit 10, ensuring that the generation of the two control signals is more in line with the actual operating state of the motor, thereby improving the accuracy of the signal gating control unit 30 in selecting the target control signal, and ultimately ensuring the control accuracy and response speed of the motor drive control system, so that it can better adapt to the complex and ever-changing motion conditions of the electric drive joints of the humanoid robot.

[0032] Furthermore, the signal gating control unit 30 in this embodiment is configured such that: when the motor unit 40 starts running, the signal gating control unit 30 selects the first control signal as the target control signal; when the motor unit 40 reaches a stable operating state and the second control signal converges, the signal gating control unit 30 selects the second control signal as the target control signal.

[0033] In this embodiment, when the motor unit 40 starts running, the signal gating control unit 30 selects the first control signal from the first control unit 20 to control the motor unit 40. This ensures that the motor unit 40 operates using the traditional motor vector control algorithm during the initial stage of operation, enabling stable and reliable operation during startup. While the motor unit 40 operates using the vector control algorithm, the second control unit 10 simultaneously generates a second control signal and sends it to the signal gating control unit 30. When the motor unit 40 reaches a stable operating state under the control of the first control signal and the second control signal converges, the signal gating control unit 30 selects the second control signal to control the motor unit 40 to continue operating. This ensures safe startup and operation of the motor unit 40 while improving the adaptability of the motor drive control system.

[0034] Understandably, in this embodiment, the stable operating state of the motor unit 40 can be determined by the fluctuation range of operating parameters, so that the motor can adapt to the corresponding operating conditions. The fluctuation range of operating parameters is determined by the fluctuation amplitude of motor phase current, DC bus voltage, motor speed, and motor position. By determining that the fluctuation of each operating parameter does not exceed a preset threshold within a preset time period, for example, within 500ms-1s, the fluctuation of motor phase current is ≤±3% of rated current, the fluctuation of motor speed is ≤±2% of target speed, and the deviation of the corresponding joint position of the motor is ≤±0.5°.

[0035] Understandably, in this embodiment, the second control signal is generated by the second control unit 10 running a reinforcement learning algorithm. The second control unit 10 trains a model based on the operating feedback signal of the motor unit 40 using the reinforcement learning algorithm to construct a reinforcement learning model. The convergence of the second control signal indicates that the output of the reinforcement learning model tends to stabilize and no longer changes significantly with the iteration process. It can be set to ensure that the deviation between the continuous output values ​​of the second control signal continuously satisfies the absolute value of the deviation ≤ a set convergence threshold within a preset time period of stable motor operation. For example, the convergence threshold can be set to ±1% to ±2% of the amplitude of the second control signal. After the second control signal converges, the reward function value of the reinforcement learning algorithm tends to stabilize. The reinforcement learning model obtains rewards through interaction with the motor operating environment. When the reward function value fluctuates very little and tends to the maximum value in multiple consecutive iterations, it indicates that the reinforcement learning model has learned the optimal control strategy adapted to the current operating condition.

[0036] Furthermore, the motor drive control system of this application embodiment also includes a torque shutdown module 60, which is electrically connected to the first control unit 20 and the signal gating control unit 30 respectively. The torque shutdown module 60 is configured to cut off the output torque of the motor unit 40 according to the torque shutdown signal issued by the first control unit 20 and / or the signal gating control unit 30.

[0037] Specifically, such as Figure 2 As shown, the torque shutdown module 60 of this application embodiment includes a torque shutdown logic circuit 61. When it is necessary to cut off the output torque of the motor unit 40, the first control unit 20 and the signal gating control unit 30 respectively generate torque shutdown signals. After receiving the torque shutdown signals issued by the first control unit 20 and the signal gating control unit 30, the torque shutdown logic circuit 61 selects one or both to control the output torque of the motor unit 40 to be cut off. When the torque shutdown signals of the first control unit 20 and the signal gating control unit 30 are selected at the same time, the physical mechanisms of the two hardware circuits of the first control unit 20 and the signal gating control unit 30 are different, which can realize the torque shutdown of the motor unit 40 based on dual redundancy signals, improve the safety of motor operation control, and significantly improve the safety level of the electric drive joint using the motor drive control system.

[0038] Furthermore, such as Figure 2As shown, the motor drive control system of this application embodiment also includes a brake locking module 70. The brake locking module 70 includes a brake locking logic circuit 71 and a locking mechanism 72. The locking mechanism 72 is disposed on the motor unit 40. The brake locking logic circuit 71 is electrically connected to the first control unit 20, the signal gating control unit 30 and the locking mechanism 72 respectively. The brake locking logic circuit 71 is configured to control the locking mechanism 72 to lock the motor unit 40 according to the brake locking signal issued by the first control unit 20 and / or the signal gating control unit 30.

[0039] Understandably, in this embodiment, when it is necessary to lock the motor in the motor unit 40, the first control unit 20 and the signal gating control unit 30 generate brake locking signals based on different mechanisms. After receiving the brake locking signals from the first control unit 20 and the signal gating control unit 30, the brake locking logic circuit 71 selects one or both to control the locking mechanism 72 to lock the motor in the motor unit 40. When the brake locking signals from the first control unit 20 and the signal gating control unit 30 are selected simultaneously, the physical mechanisms of the two hardware circuits of the first control unit 20 and the signal gating control unit 30 are different, which can realize brake locking control of the motor unit 40 based on dual redundancy signals, improve the safety of motor operation control, and significantly improve the safety level of the electric drive joint using the motor drive control system.

[0040] Preferably, such as Figure 2 As shown, the first control unit 20 in this embodiment includes at least one of an ARM chip, an MCU chip, and a DSP chip. In this embodiment, by applying an ARM chip, an MCU chip, or a DSP chip to the first control unit 20, the operational requirements of the vector control algorithm can be specifically matched, achieving high-precision, low-latency motor control signal calculation. The DSP chip possesses powerful digital signal processing capabilities, enabling rapid completion of core calculations such as coordinate transformation and PID adjustment in vector control, ensuring precise control of motor phase current and speed. The ARM chip, with its multi-core architecture and rich peripheral interfaces, can handle both control algorithm execution and system communication and status monitoring functions. The MCU chip can meet the basic vector control requirements in miniaturized, low-power scenarios. Through the flexible selection of ARM, MCU, and DSP chips, it adapts to humanoid robot electric drive joint control scenarios of varying complexity, while ensuring the generation efficiency and stability of the first control signal, providing reliable hardware support for smooth operation during motor startup.

[0041] Preferably, such as Figure 2As shown, the second control unit 10 in this embodiment includes an AI computing chip 11. The AI ​​computing chip can efficiently complete complex calculations such as parameter iteration and state value evaluation of the policy network in the reinforcement learning model, significantly improving the generation speed and convergence efficiency of the second control signal. Compared with traditional general-purpose chips, its dedicated neural network acceleration architecture can significantly reduce the algorithm runtime latency and meet the response requirements of real-time motor control. At the same time, the powerful computing power of the AI ​​computing chip supports the reinforcement learning model to quickly learn and adapt to different loads and working conditions of the humanoid robot joints, further enhancing the generalization ability of the control system and ensuring that the second control signal can accurately match diverse humanoid robot motion scenarios.

[0042] Preferably, such as Figure 2 As shown, the signal gating control unit 30 in this embodiment includes an FPGA chip (programmable gate array). The FPGA chip can directly implement the judgment and gating logic of the control signal through hardware logic circuits, without the need for operating system scheduling. Compared with the software implementation, it significantly shortens the response time of signal switching and meets the real-time requirements of motor control. Its parallel processing architecture can simultaneously receive the control signals of the first control unit 20 and the second control unit 10, as well as the operation feedback signal of the data acquisition module 50, and complete the comparison and judgment of multiple parameters in parallel, avoiding signal congestion or delay. At the same time, the programmable characteristics of the FPGA chip support the flexible adjustment of the gating logic according to the control requirements of different humanoid robot joints. It can adapt to diverse control strategies without changing the hardware, improving the flexibility and scalability of the control system.

[0043] Furthermore, such as Figure 2 As shown, the first control module in this embodiment further includes a drive circuit 80 and a power device 90. The FPGA chip is electrically connected to the motor unit 40 via the drive circuit 80 and the power device 90. The drive circuit 80 is also electrically connected to the torque shutdown logic circuit 61. The drive circuit 80 receives the torque shutdown control signal from the torque shutdown logic circuit 61, for example, cutting off the power supply to the drive circuit, thereby turning off the output of the drive circuit to achieve the torque shutdown effect.

[0044] Specifically, the drive circuit 80 amplifies, isolates, and converts the format of the first and second control signals to provide the power device 90 with drive signals that meet its operating requirements. Specifically, the drive circuit 80 receives low-amplitude, low-power control commands (such as torque commands and PWM drive signals corresponding to current setpoints) output by the signal selection control unit 30, increases the signal power through an internal amplification circuit, and simultaneously prevents high-voltage and high-current interference from the power circuit from entering the control circuit through an isolation circuit, ensuring the stable operation of the entire system.

[0045] Specifically, the power device 90 is used to control the switching of the circuit and the transmission of electrical energy according to the drive signal output by the drive circuit 80, converting the input DC power into AC power required by the motor unit 40 or adjusting the power parameters. In specific applications, the power device 90 forms an inverter circuit, such as MOSFETs, which, under the control of the drive signal, realizes the conversion of DC bus voltage to three-phase AC power of the motor stator through high-frequency switching, while precisely adjusting the amplitude, frequency, and phase of the output AC power to match the torque and speed requirements corresponding to the control signal. In addition, the power device 90 can handle large currents and high voltages in the circuit, and its switching characteristics directly affect the response speed and efficiency of the motor drive.

[0046] In this embodiment, the modularization of the system circuit is achieved through the first control module, the second control unit 10, the torque shutdown module 60, and the brake locking module 70. This provides a reliable hardware circuit platform for the humanoid robot's electric drive joints to autonomously adapt to different humanoid robots, and for real-time optimization of the electric drive joints and system performance and efficiency. It also realizes highly reliable torque shutdown and brake locking functions, as well as safe and reliable intelligent control of the humanoid robot's electric drive joints.

[0047] This application embodiment also provides an electrically driven joint structure, which includes the motor drive control system and joint module in the above embodiment, wherein the motor unit 40 in the motor drive control system and the joint module are drivenly connected.

[0048] Understandably, the electrically driven joint structure in this application embodiment is mainly applied to the movable joint parts of humanoid robots, such as the waist, shoulder, hip, hand, wrist, and ankle joints of humanoid robots, and can also be extended to the joints of robot dogs, robotic arms, etc.

[0049] like Figure 3 As shown, this application embodiment also provides a control method for an electrically driven joint. This control method is applicable to the electrically driven joint structure in the above embodiments, and the control method includes: Step S1: After the electric drive joint is installed on the predetermined device, the debugging mode is started. The first control unit 20 is controlled to generate a first control signal based on the vector control algorithm and the operation feedback signal of the motor unit 40. The second control unit 10 is controlled to run in parallel with the first control unit 20 and generate a second control signal based on the reinforcement learning algorithm and the operation feedback signal of the motor unit 40. Step S2: The signal gating control unit 30 receives the first control signal, the second control signal, and the operation feedback signal of the motor unit 40. It controls the operation of the motor unit 40 through the first control signal and at the same time determines whether the second control signal has converged. Step S3: If the second control signal does not converge or the electric drive joint does not reach a stable operating state, the motor unit 40 continues to be controlled by the first control signal; if the second control signal converges and the electric drive joint reaches a stable operating state, the signal gating control unit 30 controls the motor unit 40 to operate by the second control signal. Step S4: After the electric drive joint reaches a stable operating state under the control of the second control signal, the debugging mode is turned off. Based on the first and second control signals, the motor unit 40 is controlled to run in a preset operating mode to make the electric drive joint move.

[0050] Understandably, the predetermined device in this application embodiment includes a humanoid robot. When the electric drive joint is applied to different humanoid robots, due to the differences in the structure, degrees of freedom, load characteristics, and motion conditions of different humanoid robots, the electric drive joint needs to be debugged after it is installed on the humanoid robot. The debugging mode is started first. In the debugging mode, the first control unit 20 and the second control unit 10 are controlled to run in parallel. The signal selection control unit 30 receives the first control signal sent by the first control unit 20, the second control signal sent by the second control unit 10, and the running feedback signal sent by the data acquisition module 50. The first control signal is used as the target control signal to control the motor unit 40 to run. During the operation of the motor unit 40, the real-time running feedback signal is used to detect whether the electric drive joint has reached a stable operating state, and at the same time, it is detected whether the second control signal has converged. When the electric drive joint has reached a stable operating state and the second control signal has converged, the signal selection control unit 30 selects the second control signal sent by the second control unit 10 as the target control signal to control the motor unit 40 to run. When the motor unit 40 reaches a stable operating state again via the second control signal, the debugging mode is turned off and the motor unit 40 is controlled to run in a preset operating mode to drive the joint modules of the electrically driven joints. This enables the humanoid robot to perform dynamic load conditions such as walking and climbing stairs, improving control response speed and operational stability. In this way, the humanoid robot can ensure basic stable operation during the initial joint self-matching stage, while optimizing the reinforcement learning model in the second control unit 10 to serve as the intelligent motor control model.

[0051] Furthermore, the preset operating modes in this application embodiment include a first operating mode and a second operating mode.

[0052] In the first operating mode, the second control signal is configured as the target control signal, and the first control signal is used as a monitoring and comparison signal to monitor and adjust the second control signal in real time. Thus, in the first operating mode, the second control signal drives the motor unit 40, while the first control signal is used as a monitoring and comparison signal to verify the rationality of the second control signal in real time and dynamically correct it. This serves as a safety monitoring or redundancy signal for the second control signal, achieving both intelligence and safety, while also improving the joint movement accuracy of the humanoid robot.

[0053] In the second operating mode, the first control signal is used as the target control signal. Simultaneously, the second control signal is used to calculate the control parameter correction amount corresponding to load changes or inertia changes in the electrically driven joints. Based on the control parameter correction amount, the first control signal is optimized. Thus, in the second operating mode, the motor unit 40 is driven by the first control signal as the target control signal, while the second control signal optimizes the changes in the first control signal caused by load changes or inertia changes. The second control signal can self-match the control changes caused by load changes or robot system inertia changes, significantly improving the adaptability of the humanoid robot's joints to load changes.

[0054] Furthermore, the control method in this application embodiment also includes: When torque shutdown of the electrically driven joint is required, the torque shutdown module 60 receives torque shutdown signals synchronously issued by the first control unit 20 and the signal gating control unit 30, and cuts off the output torque between the motor unit 40 and the joint module based on the torque shutdown signal. Understandably, when the humanoid robot's electrically driven joint moves under the action of the motor unit 40, the data acquisition module 50 collects the operational feedback signals of the electrically driven joint in real time. When torque shutdown of the electrically driven joint is required, the first control unit 20 and the signal gating control unit 30 synchronously generate a torque shutdown signal, for example, a high-level signal with an amplitude of 3.3V can be generated as the torque shutdown signal. After receiving the torque shutdown signal, the torque shutdown module 60's built-in torque shutdown logic circuit 61 immediately triggers the internal switching circuit to cut off the power transmission path between the motor unit 40 and the joint module, for example, cutting off the power supply to the drive circuit 80. The joint stops power output, effectively preventing potential harm to humans when the humanoid robot interacts with them.

[0055] Furthermore, the control method in this application embodiment also includes: When it is necessary to brake and lock the electrically driven joint, the brake locking module 70 receives brake locking signals synchronously issued by the first control unit 20 and the signal gating control unit 30, and drives the locking mechanism 72 to brake and lock the electrically driven joint based on the brake locking signals. For example, after the humanoid robot completes the walking task, the first control unit 20 first adjusts the electrically driven joint to the preset locking position based on the vector control algorithm. After the joint position is stable, the first control unit 20 and the signal gating control unit 30 synchronously generate brake locking signals, such as pulse signals. After receiving the two synchronous brake locking signals, the brake locking module 70 outputs a drive current to the locking mechanism 72. After the locking mechanism 72 is energized, it realizes mechanical locking of the electrically driven joint through its internal electromagnetic structure. After locking, the brake locking module 70 sends a real-time locking signal to the first control unit 20 and the signal selection control unit 30 to confirm that the locking state is valid. It can reliably lock the brake in scenarios such as when the humanoid robot is standing still or when there is a sudden power outage, preventing the humanoid robot from tipping over due to accidental rotation and improving the safety of the humanoid robot.

[0056] Furthermore, the operation feedback signal of the motor unit 40 in this embodiment includes at least the motor phase current, DC bus voltage, motor speed, and motor position.

[0057] In this embodiment of the application, the step of generating the first control signal based on the vector control algorithm combined with the operation feedback signal of the motor unit 40 includes: constructing a motor closed-loop control model using the motor phase current, DC bus voltage, motor speed and motor position as control measurement signals, and generating the first control signal through the motor closed-loop control model.

[0058] like Figure 4 As shown, the vector control algorithm in this embodiment adopts a three-loop control architecture consisting of a position control loop, a speed control loop, and a current control loop. All three loops use PI control, converting the three-phase AC power into independent DC control along the d-axis and q-axis through coordinate transformation. The d-axis represents the rotor flux linkage direction, and the q-axis represents the torque direction, achieving precise control of the motor torque and speed. This indicates the target position of the motor, which serves as the input to the outermost position control loop; The target motor speed is indicated by the position control loop and is used as the input to the speed control loop. The target current on the q-axis is output by the speed control loop and serves as the q-axis input to the current control loop. The q-axis current corresponds to the electromagnetic torque. This represents the target current on the d-axis, which serves as the d-axis input to the current control loop. Typically, permanent magnet synchronous motors (surface-mounted) use id=0 control, meaning the desired d-axis current is 0. Indicates the actual position of the motor. Indicates the actual speed of the motor, ( () represents the actual three-phase current of the motor. , 、( () is the operation feedback signal; This represents the DC bus voltage, which is used as the supply voltage. The position control loop is... →PI controller→output To convert position deviation into speed command; the speed control loop is →PI controller→output To convert the speed deviation into a q-axis current command; the current control loop first needs to perform coordinate transformation, including Clarke T transformation and Park T transformation. The Clarke T transformation converts the three-phase phase current ( Converted to current in α / β stationary coordinate system ( This achieves a three-dimensional to two-dimensional transformation, eliminating three-phase coupling. The Park-T transform will ( Combined with the actual position of the motor Converted to d / q rotating coordinate system current ( This achieves current decoupling. Current regulation includes two processes, one of which is... →PI controller→output One is the q-axis voltage command, and the other is... →PI controller→output This refers to the d-axis voltage command. The current control loop also includes the Park inverse transform (…). ), mainly , Combination Converted to voltage in α / β coordinate system , The Park inverse transform also uses a voltage decoupler to eliminate coupling interference between the d-axis and q-axis voltages. This is further achieved through a space vector pulse width modulator. , The signal is converted into 6 PWM signals, namely PWM1, PWM2, PWM3, PWM4, PWM5, and PWM6. These 6 PWM signals drive the three-phase inverter to output three-phase voltage and control the motor operation.

[0059] Furthermore, in this embodiment of the application, the step of generating a second control signal based on the reinforcement learning algorithm and the operation feedback signal of the motor unit 40 includes: using motor speed, motor position, motor phase current, and DC bus voltage as state parameters, using motor control commands as action parameters, constructing a reinforcement learning model by maximizing a preset reward function to learn the optimal control strategy, and outputting the second control signal through the reinforcement learning model.

[0060] like Figure 5As shown, the reinforcement learning algorithm in this embodiment interacts with the environment through an agent. The agent contains a decision-making unit of a DNN (deep neural network) and outputs a control strategy through the network. The environment is represented by the motor and load system, i.e., the humanoid robot in this embodiment. This is achieved through state... )-action( )-award( ) Iterative learning of optimal control strategy State This represents the system state perceived by the intelligent agent, including the actual speed of the motor. Actual location Three-phase actual phase current ( DC bus voltage And the deviation between the control target and the actual value (e.g.) ) etc.; actions ( ) represents the control commands output by the intelligent agent, namely the inverter's six pulse width modulation signals PWM1, PWM2, PWM3, PWM4, PWM5, and PWM6, corresponding to different switching states; reward ( The indicators used to evaluate the control effect can be designed manually, such as the actual rotational speed. near When the current or voltage exceeds the safe range, a positive reward is given; when the current or voltage exceeds the safe range, a negative reward is given. (Strategy) ) represents the decision rules of the agent, fitted by a DNN, with the input state (State) ), output action ( The three-phase inverter receives PWM signals and outputs three-phase voltage to drive the motor. The specific control process of the reinforcement learning algorithm is as follows: the agent acquires the current state (including the deviation between the desired target and the actual feedback, operational feedback signals, etc.), the agent outputs the PWM signals corresponding to each action through the optimal strategy, acts on the environment (motor), drives the motor to run, and the environment generates new observations. The environment feeds back information to the agent (i.e., the actual value of the state). The environment calculates the reward based on the control effect and feeds it back to the agent. The agent updates its strategy and optimizes the DNN parameters with the goal of "maximizing the cumulative reward". The above loop is repeated until the agent learns the optimal control strategy and achieves precise control of the motors of the electrically driven joints of the humanoid robot.

[0061] As can be seen from the above description, the embodiments of the present invention achieve the following technical effects: This solution integrates AI high-performance chips, ARM / MCU / DSP chips, FPGA chips, joint torque shutdown modules, and brake locking modules, employing a parallel redundant control architecture to achieve intelligent control of the electric drive joints of humanoid robots. While possessing strong generalization capabilities, it ensures the safety of the humanoid robot system operation, achieving safe, reliable, and highly generalizable intelligent control of the electric drive joints. It achieves the effect of autonomously adapting to the operating conditions of new humanoid robots when the electric drive joints are installed, requiring no manual debugging, and meeting the high safety requirements of humanoid robot application scenarios during operation.

[0062] For ease of description, spatial relative terms such as "above," "on top of," "on the upper surface of," "above," etc., are used herein to describe the spatial positional relationship of a device or feature as shown in the figures to other devices or features. It should be understood that spatial relative terms are intended to encompass different orientations in use or operation beyond the orientation of the device as described in the figures. For example, if the device in the figures were inverted, a device described as "above" or "on top of" other devices or structures would subsequently be positioned as "below" or "under" other devices or structures. Thus, the exemplary term "above" can include both "above" and "below." The device may also be positioned in other different ways (rotated 90 degrees or in other orientations), and the spatial relative descriptions used herein will be interpreted accordingly.

[0063] Furthermore, it should be noted that the use of terms such as "first" and "second" to define components is merely for the purpose of distinguishing the corresponding components. Unless otherwise stated, the above terms have no special meaning and therefore should not be construed as limiting the scope of protection of this invention.

[0064] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A motor drive control system, characterized in that, include: The first control module includes a first control unit (20), a signal gating control unit (30), and a motor unit (40) connected in sequence. The first control unit (20) is configured to generate a first control signal based on the operation feedback signal of the motor unit (40) through a vector control algorithm. The second control unit (10) is electrically connected to the signal gating control unit (30). The second control unit (10) is configured to operate in parallel with the first control unit (20) and generate a second control signal through a reinforcement learning algorithm based on the operation feedback signal of the motor unit (40). The signal selection control unit (30) is configured to select either the first control signal or the second control signal as the target control signal based on the operation feedback signal of the motor unit (40) and the second control signal, and control the operation of the motor unit (40) based on the target control signal.

2. The motor drive control system according to claim 1, characterized in that, The signal gating control unit (30) is configured such that when the motor unit (40) starts running, the signal gating control unit (30) selects the first control signal as the target control signal, and when the motor unit (40) reaches a stable operating state and the second control signal converges, the signal gating control unit (30) selects the second control signal as the target control signal.

3. The motor drive control system according to claim 1, characterized in that, The motor drive control system further includes a torque cut-off module (60), which is electrically connected to the first control unit (20) and the signal gating control unit (30) respectively. The torque cut-off module (60) is configured to cut off the output torque of the motor unit (40) according to the torque cut-off signal issued by the first control unit (20) and / or the signal gating control unit (30).

4. The motor drive control system according to claim 1, characterized in that, The motor drive control system further includes a brake locking module (70), which includes a brake locking logic circuit (71) and a locking mechanism (72). The locking mechanism (72) is disposed on the motor unit (40). The brake locking logic circuit (71) is electrically connected to the first control unit (20), the signal gating control unit (30), and the locking mechanism (72). The brake locking logic circuit (71) is configured to control the locking mechanism (72) to lock the motor unit (40) according to the brake locking signal issued by the first control unit (20) and / or the signal gating control unit (30).

5. The motor drive control system according to any one of claims 1 to 4, characterized in that, The first control unit (20) includes at least one of an ARM chip, an MCU chip, and a DSP chip; and / or, The second control unit (10) includes an AI computing chip (11); and / or, The signal gating control unit (30) includes an FPGA chip.

6. An electrically driven joint structure, characterized in that, The electric drive joint structure includes a motor drive control system and a joint module as described in any one of claims 1 to 5, wherein the motor unit (40) in the motor drive control system and the joint module are drivenly connected.

7. A control method for an electrically driven joint, characterized in that, The control method applicable to the electrically driven joint structure of claim 6 includes: Step S1: After the electric drive joint is installed on the predetermined device, the debugging mode is started. The first control unit (20) is controlled to generate a first control signal based on the vector control algorithm and the running feedback signal of the motor unit (40). The second control unit (10) is controlled to run in parallel with the first control unit (20) and generate a second control signal based on the reinforcement learning algorithm and the running feedback signal of the motor unit (40). Step S2: The signal gating control unit (30) receives the first control signal, the second control signal and the operation feedback signal of the motor unit (40), controls the operation of the motor unit (40) through the first control signal, and at the same time determines whether the second control signal converges; Step S3: If the second control signal does not converge or the electric drive joint does not reach a stable operating state, the motor unit (40) continues to be controlled by the first control signal; if the second control signal converges and the electric drive joint reaches a stable operating state, the signal gating control unit (30) controls the motor unit (40) to operate by the second control signal. Step S4: After the electric drive joint reaches a stable operating state under the control of the second control signal, the debugging mode is turned off. Based on the first control signal and the second control signal, the motor unit (40) is controlled to operate in a preset operating mode to make the electric drive joint move.

8. The control method for an electrically driven joint according to claim 7, characterized in that, The preset operating modes include: A first operating mode is configured to use the second control signal as the target control signal and use the first control signal as a monitoring comparison signal to monitor and adjust the second control signal in real time. The second operating mode is configured to use the first control signal as the target control signal, and simultaneously calculate the control parameter correction amount corresponding to the load change or inertia change of the electric drive joint through the second control signal, and optimize the first control signal based on the control parameter correction amount.

9. The control method for an electrically driven joint according to claim 7, characterized in that, The control method further includes: When torque shutdown of the electrically driven joint is required, the torque shutdown module (60) receives a torque shutdown signal synchronously issued by the first control unit (20) and the signal gating control unit (30), and cuts off the output torque between the motor unit (40) and the joint module based on the torque shutdown signal; and / or, When it is necessary to brake the electrically driven joint, the brake locking module (70) receives the brake locking signal synchronously issued by the first control unit (20) and the signal gating control unit (30), and drives the locking mechanism (72) to brake the electrically driven joint based on the brake locking signal.

10. The control method for an electrically driven joint according to claim 7, characterized in that, The operating feedback signal of the motor unit (40) includes at least one of the following: motor phase current, DC bus voltage, motor speed, and motor position: The step of generating the first control signal based on the vector control algorithm combined with the operating feedback signal of the motor unit (40) includes: constructing a motor closed-loop control model using the motor phase current, the DC bus voltage, the motor speed, and the motor position as control measurement signals, and generating the first control signal through the motor closed-loop control model; and / or, The step of generating the second control signal based on the reinforcement learning algorithm and the operation feedback signal of the motor unit (40) includes: using the motor speed, the motor phase current, and the DC bus voltage as state parameters, using the motor control command as action parameters, constructing a reinforcement learning model by maximizing the preset reward function to learn the optimal control strategy, and outputting the second control signal through the reinforcement learning model.

Images