Servo system optimization method and system based on algorithm control

Abstract

The invention provides a servo system optimization method and system based on algorithm control, and relates to a motor control technology, and the method comprises the steps: firstly obtaining a real-time dynamic response parameter set during the dynamic operation of a servo device, covering inertia delay, torque fluctuation and position deviation parameters, and a corresponding optimization demand set, including optimization demands for each parameter; performing multi-level control signal generation on the real-time dynamic response parameter set according to the optimization demand priority, generating a control signal sequence set matched with a first compensation signal, a second compensation signal, a third compensation signal and the like, inputting the control signal sequence set into an execution mechanism, and driving the execution mechanism to dynamically respond according to an updating track, and finally, the control signal sequence set is optimized according to a feedback signal set generated by an execution mechanism, so that operation parameters of the servo device can be comprehensively considered, control signals are generated in a targeted manner, dynamic adjustment and optimization are realized, and the performance of the servo system is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of motor control, and in particular to a servo system optimization method and system based on algorithm control. Background Art

[0002] In the field of industrial automation and precision control, servo systems, as key actuators, are widely used in various high-precision motion control scenarios, such as CNC machine tool processing, robot operation, aerospace equipment manufacturing, etc. The quality of their performance directly affects product quality, production efficiency, and equipment stability and reliability.

[0003] Existing servo system control methods lack systematicity and pertinence, often adopting fixed control strategies without considering the differences in optimization requirements of different parameters, and failing to flexibly adjust control signals according to actual conditions. In terms of driving the motion of actuators, existing technologies make it difficult to achieve real-time dynamic adjustment of the motion trajectory of actuators. Actuators usually operate according to a preset fixed trajectory. When encountering external interference or changes in working conditions, they are unable to update the motion trajectory in time according to actual needs, resulting in reduced operating accuracy and failure to meet complex and changing work task requirements.

[0004] Moreover, most existing control methods use open-loop control or simple feedback control. The open-loop control method cannot adjust the control parameters according to the actual operation of the system. Once interference or parameter changes occur, the system performance will drop sharply. Although simple feedback control can correct system deviations to a certain extent, it lacks a comprehensive optimization mechanism and cannot deeply optimize the control signal according to the comprehensive influence of different parameters, making it difficult to ensure that the servo system is always in the best operating state. Summary of the invention

[0005] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a servo system optimization method based on algorithm control, the method comprising:

[0006] Acquire a real-time dynamic response parameter set generated by the servo device during dynamic operation, and an optimization requirement set corresponding to the servo device; wherein the real-time dynamic response parameter set includes an inertia delay parameter, a torque fluctuation parameter, and a position deviation parameter generated by the servo device when executing a target motion trajectory, and the optimization requirement set includes a first optimization requirement for the inertia delay parameter, a second optimization requirement for the torque fluctuation parameter, and a third optimization requirement for the position deviation parameter;

[0007] According to the demand priority corresponding to each optimization demand in the optimization demand set, multi-level control signal generation is performed on the real-time dynamic response parameter set to obtain a control signal sequence set matching each optimization demand; wherein the multi-level control signal generation includes generating a first compensation signal for the inertia delay parameter, generating a second compensation signal for the torque fluctuation parameter, and generating a third compensation signal for the position deviation parameter;

[0008] Inputting the control signal sequence set to the actuator of the servo device to drive the actuator to dynamically respond according to the updated motion trajectory;

[0009] The control signal sequence set is optimized based on the feedback signal set generated by the actuator during the dynamic response process.

[0010] On the other hand, an embodiment of the present invention also provides a servo system optimization system based on algorithm control, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0011] Based on the above aspects, the embodiment of the present application can generate matching control signals in a targeted manner according to the importance of different parameters by accurately obtaining the real-time dynamic response parameter set and the corresponding optimization requirement set generated by the servo device during dynamic operation, and operating according to the requirement priority corresponding to each optimization requirement. Specifically, the first compensation signal, the second compensation signal and the third compensation signal are generated for the inertia delay parameter, the torque fluctuation parameter and the position deviation parameter, respectively, which effectively and accurately solves the problems caused by different parameters, and greatly improves the flexibility and effectiveness of control compared with the traditional method. Traditional methods often adopt a unified control method, which cannot be customized according to the characteristics and priorities of different parameters, while the multi-level control signal generation method of this method can more finely adjust the operation of the servo system. Then, the control signal sequence set is input into the actuator of the servo device, driving it to respond dynamically according to the updated motion trajectory, realizing the precise control and dynamic adjustment of the motion trajectory of the servo device, so that the servo device can quickly adapt to various complex working scenes and task requirements, greatly expanding the application scope of the servo system. Finally, by optimizing the control signal sequence set based on the feedback signal set generated by the actuator during the dynamic response process, the control signal can be continuously optimized according to the actual operating conditions, continuously improving the performance of the servo system and ensuring that it is always in the best operating state. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1It is a schematic diagram of the execution flow of the servo system optimization method based on algorithm control provided by an embodiment of the present invention.

[0013] Figure 2 It is a schematic diagram of the hardware architecture of a servo system optimization system based on algorithm control provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0014] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 1 is a flow chart of a servo system optimization method based on algorithm control provided by an embodiment of the present invention. The servo system optimization method based on algorithm control is introduced in detail below.

[0015] Step S110, obtaining a real-time dynamic response parameter set generated by the servo device during dynamic operation, and an optimization requirement set corresponding to the servo device. The real-time dynamic response parameter set includes an inertia delay parameter, a torque fluctuation parameter, and a position deviation parameter generated by the servo device when executing the target motion trajectory, and the optimization requirement set includes a first optimization requirement for the inertia delay parameter, a second optimization requirement for the torque fluctuation parameter, and a third optimization requirement for the position deviation parameter.

[0016] Taking a quadruped robot as an example, during the operation of the quadruped robot, the movement modes of the quadruped robot are complex and diverse, such as walking, running, crossing obstacles, etc. The above actions require servo devices to accurately control the movement of each joint.

[0017] When the quadruped robot executes a target motion trajectory (such as crossing an obstacle of a specific height and width), the servo device generates real-time dynamic response parameters.

[0018] For example, in terms of inertia delay parameters, the leg joints of a quadruped robot are driven by a servo device. When the legs need to be lifted quickly to cross an obstacle, inertia delay will occur due to the inertia of the mechanical structure itself and the response characteristics of the servo device. For example, when the robot's control instructions require the leg joints to accelerate from a stationary state to a specific angular velocity within 0.1 seconds in order to be lifted smoothly, but due to inertia delay, it may actually take 0.15 seconds to achieve this. The 0.05-second delay is part of the inertia delay parameter.

[0019] For another example, in terms of torque fluctuation parameters, when a quadruped robot walks, different terrains will impose different loads on the leg joints. When the robot walks from flat ground to a slope, the leg joints require greater torque to support the weight of the robot and push it forward. In order to adapt to this load change, the servo device will adjust the output torque, but torque fluctuations may occur in this process. For example, when walking on normal flat ground, the torque of the leg joints is stable near a value, but when walking on a slope, the torque may fluctuate up and down from this stable value in a short period of time, and the fluctuation range may reach ±10%. This fluctuation value is the torque fluctuation parameter.

[0020] For another example, in terms of position deviation parameters, each joint of a quadruped robot has its predetermined motion position to ensure the overall balance and stable motion of the robot. However, in actual motion, position deviation may occur due to various factors. For example, when the robot turns, due to the uneven distribution of inertia and ground friction, the joint of a leg may not reach the theoretically accurate position, and the deviation may reach several millimeters. This deviation is the position deviation parameter.

[0021] At the same time, there is a corresponding set of optimization requirements for these dynamic response parameters. For example, for the first optimization requirement (for the inertia delay parameter): in the fast movement scenario of the quadruped robot, such as running, it is required to minimize the inertia delay. Because a large inertia delay may cause the robot's movements to be uncoordinated, affecting the speed and stability of the running. For example, if the robot wants to change direction quickly, the inertia delay of the leg joints is too large, and it will not be able to respond to the control command in time, which may cause the robot to fall.

[0022] For another example, regarding the second optimization requirement (for the torque fluctuation parameter): when a quadruped robot crosses different terrains (such as from hard ground to soft sand), a stable torque output is required to ensure that the robot's joints are not damaged due to excessive torque fluctuations and that it can move smoothly on different terrains.

[0023] For another example, regarding the third optimization requirement (position deviation parameter): when the quadruped robot performs precise movements (such as standing still or performing delicate operations), the position deviation is required to be controlled within an extremely small range to ensure the overall balance of the robot and the accuracy of the movement.

[0024] Step S120, according to the demand priority corresponding to each optimization demand in the optimization demand set, multi-level control signal generation is performed on the real-time dynamic response parameter set to obtain a control signal sequence set matching each optimization demand. The multi-level control signal generation includes generating a first compensation signal for the inertia delay parameter, generating a second compensation signal for the torque fluctuation parameter, and generating a third compensation signal for the position deviation parameter.

[0025] In this embodiment, the optimization requirements of the servo device of the quadruped robot have different priorities. Assuming that when the robot performs a task that crosses complex terrain (including slopes, potholes, etc.) and requires fast movement, the first optimization requirement (reducing inertia delay) has the highest priority, because fast response can ensure that the robot moves stably and quickly on complex terrain; the second optimization requirement (stabilizing torque fluctuations) has the second highest priority, mainly to protect the joints and ensure the smoothness of movement; the third optimization requirement (reducing position deviation) has a relatively low priority, because in such a complex and fast-moving scenario, the position deviation is acceptable within a certain range.

[0026] Generation of the first compensation signal for the inertia delay parameter: Determine the direction and size of the compensation based on the priority of the first optimization requirement and the actual situation of the inertia delay parameter. For example, if the inertia delay causes the leg joint response time to be extended by 0.05 seconds, and the acceptable delay range (based on the robot design and task requirements) is no more than 0.03 seconds, then a compensation signal needs to be generated. This compensation signal should be able to accelerate the servo device's response speed to the command so that the next time the leg joint moves, the inertia delay can be reduced to an acceptable range. Some parameters in the control algorithm can be adjusted, such as increasing the pre-drive current to the motor, so that the motor can overcome inertia faster and start to rotate.

[0027] Generation of the second compensation signal for the torque fluctuation parameter: Since the torque fluctuation parameter is affected by factors such as terrain, the current terrain conditions and the robot's motion state need to be considered when determining the compensation signal. For example, when the robot enters soft sand from hard ground, the torque fluctuation is large. According to the pre-set torque fluctuation allowable range (such as ±10%), if the actual fluctuation reaches ±15%, a compensation signal needs to be generated to stabilize the torque output. The compensation signal may be achieved by adjusting the gain parameter in the torque control loop of the servo device to make the torque output smoother and reduce fluctuations.

[0028] Generation of the third compensation signal for position deviation parameters: When the position deviation exceeds the acceptable range (such as a few millimeters), a compensation signal needs to be generated. For example, when the robot is standing still, the position deviation of a joint reaches 5 mm due to uneven ground or slight deformation of its own structure, while the acceptable deviation range is 2 mm. At this time, the compensation signal may be to adjust the proportional-integral-differential (PID) parameters in the position control algorithm so that the servo device can fine-tune the joint position and reduce the deviation to an acceptable range.

[0029] Through the above operations, a control signal sequence set matching each optimization requirement is obtained.

[0030] Step S130 , inputting the control signal sequence set to the actuator of the servo device, driving the actuator to perform dynamic response according to the updated motion trajectory.

[0031] In this embodiment, the generated control signal sequence set is input into the servo device actuator of the quadruped robot.

[0032] First, the execution timestamp and execution strength value corresponding to each control signal in the control signal sequence set are parsed. For example, for the signal that controls the leg joint to lift, the execution timestamp may be 0.5 seconds after the robot starts to cross the obstacle, and the execution strength value determines the speed and strength of the joint lifting.

[0033] Then, the control signal sequence set is divided into multiple control signal subsets according to the execution timestamp. For a quadruped robot, each movement phase (such as preparing to lift the leg, lifting the leg, crossing the leg, putting the leg down, etc.) has a corresponding control signal subset. At the start of the leg lifting movement phase, the corresponding control signal subset is loaded into the signal buffer queue of the actuator. And the control signals in the signal buffer queue are power amplified according to the execution strength value. For example, if the execution strength value requires the leg joint to be lifted with greater force, then the control signal will be power amplified accordingly to ensure that the leg can overcome gravity and inertia and be lifted smoothly.

[0034] The amplified control signal is converted into a physical drive instruction through the drive circuit of the actuator. In this process, the voltage amplitude, current frequency and pulse width modulation (PWM) parameters in the amplified control signal are obtained. For example, assuming that the voltage amplitude is 12V, the current frequency is 100Hz, and the duty cycle in the PWM parameter is 50%. The target output voltage and target output current of the drive circuit are calculated based on the voltage amplitude and current frequency, and then a pulse waveform sequence matching the target output voltage and target output current is generated based on the PWM parameters. The pulse waveform sequence is input to the power switch device of the drive circuit to control the on time and off time of the power switch device. Through the alternating on and off operations of the power switch device, the pulse waveform sequence is converted into a physical drive instruction with a predetermined duty cycle and frequency, thereby adjusting the motor speed, output torque and position feedback accuracy of the actuator.

[0035] In this process, it is also necessary to monitor the real-time motion state of the actuator during the motion phase. For example, during the leg lifting process, the angle, speed and other parameters of the leg joints are monitored by sensors. If an abnormal motion state is detected, such as a sudden drop in the speed of the leg joint or an angle that exceeds the normal range, the execution of the current control signal subset will be interrupted and the abnormal recovery mechanism will be triggered. The abnormal recovery mechanism may include readjusting the control signal, reducing the motion speed or making some emergency posture adjustments to ensure the stable operation of the quadruped robot.

[0036] Step S140: optimizing the control signal sequence set based on the feedback signal set generated by the actuator in the dynamic response process.

[0037] During the movement of the quadruped robot, the control signal sequence set is optimized based on the feedback signal set generated by the actuator.

[0038] First, collect the feedback signal set generated by the actuator during the dynamic response process. The acceleration feedback signal and angular velocity feedback signal of the actuator (leg joint) are collected through the inertial measurement unit in the servo device of the quadruped robot. For example, when the leg joint is lifted up by acceleration, the inertial measurement unit can measure the changes in the acceleration and angular velocity of the joint. The real-time torque feedback signal of the actuator is collected through the torque sensor to understand the torque size of the leg joint during the movement. The actual position feedback signal of the actuator is collected through the position encoder to determine whether the leg joint has reached the predetermined position.

[0039] These feedback signals are subjected to noise filtering. For acceleration feedback signals, high-frequency vibration noise components are identified, such as high-frequency noise generated by mechanical vibration of the robot during movement, and these high-frequency vibration noise components are attenuated using a low-pass filter. For angular velocity feedback signals, instantaneous pulse interference components are detected, such as pulse interference generated when a joint is suddenly impacted by an external force, and the instantaneous pulse interference components are smoothed using a sliding window mean filter algorithm. For real-time torque feedback signals, periodic fluctuation noise is extracted, such as periodic fluctuations caused by the electromagnetic characteristics of the motor, and an adaptive notch filter is used to eliminate the periodic fluctuation noise. For actual position feedback signals, the quantization error components are analyzed, and the quantization error components are compensated and corrected using an interpolation algorithm.

[0040] The noise-filtered signal is converted into actual dynamic performance indicators. The denoised acceleration data is converted into actual inertial delay parameters, the denoised angular velocity data is combined with the position data to generate actual position deviation parameters, and the denoised torque data is converted into actual torque fluctuation parameters.

[0041] According to the degree of difference between these actual dynamic performance indicators and the preset performance threshold, each control signal in the control signal sequence set is dynamically weighted. For example, if the preset inertia delay threshold is 0.03 seconds, and the actual inertia delay reaches 0.04 seconds, the absolute value of the difference between the inertia delay deviation value and the preset inertia delay threshold is calculated to be 0.01 seconds. The weight adjustment coefficient corresponding to each control signal is determined based on the absolute value of the difference. Since the larger the absolute value of the difference, the greater the need for adjustment of the control signal, the weight adjustment coefficient is positively correlated with the absolute value of the difference. Based on the weight adjustment coefficient, the strength value of each control signal in the control signal sequence set is scaled to generate an intermediate control signal sequence.

[0042] Then, the execution timestamp of each control signal in the intermediate control signal sequence is offset compensated. The phase detection signal of the quadruped robot in the current motion cycle is obtained, and the zero-crossing point and peak point are analyzed. The dynamic response phase angle of the quadruped robot is determined according to the zero-crossing point and the peak point, and the theoretical optimal execution time of each control signal is calculated based on the dynamic response phase angle. The time offset between the original execution timestamp of each control signal in the intermediate control signal sequence and the theoretical optimal execution time is compared. If the time offset is greater than the preset offset threshold, for example, the preset offset threshold is 0.01 seconds, and the actual time offset is 0.02 seconds, then the original execution timestamp is corrected according to the theoretical optimal execution time; if the time offset is less than or equal to the preset offset threshold, the original execution timestamp is kept unchanged. The control signal sequence after intensity scaling and timestamp offset is used as the optimized control signal sequence set, and it is iteratively input into the actuator, so as to continuously optimize the motion performance of the quadruped robot.

[0043] Based on the above steps, the embodiment of the present application can generate matching control signals in a targeted manner according to the importance of different parameters by accurately obtaining the real-time dynamic response parameter set and the corresponding optimization requirement set generated by the servo device during dynamic operation, and operating according to the requirement priority corresponding to each optimization requirement. Specifically, the first compensation signal, the second compensation signal and the third compensation signal are generated for the inertia delay parameter, the torque fluctuation parameter and the position deviation parameter, respectively, which effectively and accurately solves the problems caused by different parameters, and greatly improves the flexibility and effectiveness of control compared with the traditional method. Traditional methods often adopt a unified control method, which cannot be customized according to the characteristics and priorities of different parameters, and the multi-level control signal generation method of this method can more finely adjust the operation of the servo system. Then, the control signal sequence set is input into the actuator of the servo device, driving it to respond dynamically according to the updated motion trajectory, realizing the precise control and dynamic adjustment of the motion trajectory of the servo device, so that the servo device can quickly adapt to various complex working scenes and task requirements, greatly expanding the application scope of the servo system. Finally, by optimizing the control signal sequence set based on the feedback signal set generated by the actuator during the dynamic response process, the control signal can be continuously optimized according to the actual operating conditions, continuously improving the performance of the servo system and ensuring that it is always in the best operating state.

[0044] In a possible implementation, step S140 includes:

[0045] Step S141, collecting a feedback signal set generated by the actuator during the dynamic response process, and determining the current dynamic performance index of the servo device based on the actual inertia delay parameter, the actual torque fluctuation parameter and the actual position deviation parameter in the feedback signal set.

[0046] For example, the inertial measurement unit in the servo device can accurately collect the acceleration feedback signal and angular velocity feedback signal of the actuator (leg joint) during the movement. For example, when the quadruped robot performs a rapid turning action, the leg joints will produce complex acceleration and angular velocity changes, and the inertial measurement unit will record these data in real time. The torque sensor is responsible for collecting the real-time torque feedback signal of the actuator. When the quadruped robot walks on different terrains, such as walking from flat ground to a slope with a certain slope, the torque borne by the leg joints will change significantly, and the torque sensor can accurately capture this change. The position encoder can collect the actual position feedback signal of the actuator. When the quadruped robot performs actions such as crossing obstacles or precise pace adjustment, the position encoder can give corresponding feedback on whether the leg joints have accurately reached the predetermined position.

[0047] Assume that the quadruped robot is executing a complex walking path, which includes acceleration, deceleration, turning, and crossing small obstacles. In this process, if the actual inertia delay parameter shows that the delay time from receiving the command to starting the action of the leg joint is longer than expected by a certain value, for example, the expected inertia delay should not exceed 0.03 seconds in a certain action, but the actual measurement is 0.05 seconds; the actual torque fluctuation parameter indicates that the torque fluctuation amplitude exceeds the design allowable range when switching between different terrains, such as the normal allowable torque fluctuation range is ±10%, and the actual fluctuation reaches ±15%; the actual position deviation parameter shows that the leg joint position deviates from the predetermined position in some key actions by more than the acceptable deviation, for example, the acceptable position deviation is 2 mm, but the actual deviation reaches 3 mm. The above data combined reflect that the current dynamic performance indicators of the servo device are not good, which may affect the overall motion performance of the quadruped robot, such as walking stability, movement accuracy, and adaptability to different terrains.

[0048] Step S142, dynamically adjust the weight of each control signal in the control signal sequence set according to the degree of difference between the current dynamic performance indicator and the preset performance threshold, generate an optimized control signal sequence set, and iteratively input the optimized control signal sequence set to the actuator.

[0049] Taking the inertia delay parameter as an example, if the preset inertia delay threshold is 0.03 seconds and the actual measurement is 0.05 seconds, the absolute value of the difference between the two is 0.02 seconds. The difference indicates that the control signal needs to be adjusted. The larger the difference, the larger the corresponding weight adjustment coefficient. For the torque fluctuation parameter, if the preset fluctuation range is ±10% and the actual fluctuation reaches ±15%, the weight adjustment coefficient is also determined according to the degree of exceeding the range. The same is true for the position deviation parameter. Based on these weight adjustment coefficients, the intensity value of each control signal in the control signal sequence set is scaled to generate an intermediate control signal sequence. For example, the original intensity value of a control signal is to make the leg joint move at a certain speed. If it needs to be adjusted due to a large inertia delay, the intensity value may be increased according to the weight adjustment coefficient to speed up the response speed of the joint.

[0050] When offset compensation is performed on the execution timestamp of each control signal in the intermediate control signal sequence, the phase detection signal of the quadruped robot contains important information in the current motion cycle. Then, the zero crossing points and peak points in the phase detection signal are parsed. For example, in the motion cycle of the robot leg joint, the zero crossing points and peak points correspond to the specific state transition points of the joint motion. The dynamic response phase angle of the quadruped robot is determined according to these zero crossing points and peak points, and the theoretical optimal execution time of each control signal is calculated based on the dynamic response phase angle. The time offset between the original execution timestamp and the theoretical optimal execution time of each control signal in the intermediate control signal sequence is compared. If the time offset is greater than the preset offset threshold, for example, the preset offset threshold is 0.01 seconds, and the actual time offset is 0.02 seconds, the original execution timestamp is corrected according to the theoretical optimal execution time. If the time offset is less than or equal to the preset offset threshold, the original execution timestamp is kept unchanged. After such intensity scaling and timestamp offset, the control signal sequence becomes an optimized control signal sequence set, and then the optimized control signal sequence set is iteratively input to the actuator. In this way, in the subsequent movement process, the quadruped robot can make more accurate and efficient movements according to the optimized control signals, and continuously improve its movement performance in complex environments.

[0051] In a possible implementation, each optimization requirement in the optimization requirement set includes a requirement type identifier and a requirement constraint condition, and step S120 includes:

[0052] Step S121, for each optimization requirement in the optimization requirement set, determining the target dynamic response parameter corresponding to the optimization requirement according to the requirement type identifier, wherein the requirement type identifier includes an inertia delay optimization identifier, a torque fluctuation optimization identifier, and a position deviation optimization identifier.

[0053] Step S122: extracting the constraint boundary of the target dynamic response parameter according to the parameter adjustment range defined in the demand constraint condition, and obtaining the allowable fluctuation range of the target dynamic response parameter.

[0054] Step S123, based on the allowable fluctuation range and the demand priority, construct a parameter adjustment function corresponding to the target dynamic response parameter, and call the parameter adjustment function to calculate a compensation signal for the target dynamic response parameter. The compensation signal calculation includes calculating a reverse compensation amount for a parameter deviation value that exceeds the allowable fluctuation range.

[0055] Step S124: Map the reverse compensation amount to a corresponding control signal strength value, and sort the control signal strength values ​​according to the demand priority to generate the control signal sequence set.

[0056] In a possible implementation manner, the demand constraint condition includes a time constraint condition and a space constraint condition, and step S122 includes:

[0057] Step S1221, parsing the maximum allowed delay time and the minimum allowed response time in the time constraint condition, and determining a first time constraint interval corresponding to the inertia delay parameter.

[0058] Step S1222, analyzing the maximum allowable position deviation distance and the minimum allowable torque fluctuation amplitude in the spatial constraint condition, and determining a first spatial constraint interval corresponding to the position deviation parameter and a second spatial constraint interval corresponding to the torque fluctuation parameter.

[0059] Step S1223: constructing a first allowable fluctuation interval of the inertia delay parameter, a second allowable fluctuation interval of the position deviation parameter, and a third allowable fluctuation interval of the torque fluctuation parameter, respectively, according to the first time constraint interval, the first space constraint interval, and the second space constraint interval. The first allowable fluctuation interval is used to limit the time fluctuation range of the inertia delay parameter, the second allowable fluctuation interval is used to limit the spatial offset range of the position deviation parameter, and the third allowable fluctuation interval is used to limit the amplitude variation range of the torque fluctuation parameter.

[0060] And, step S123 includes:

[0061] Step S1231, determining a compensation weight coefficient corresponding to each allowed fluctuation interval according to the demand priority, wherein the compensation weight coefficient is positively correlated with the demand priority.

[0062] Step S1232: input the upper limit value and the lower limit value of the allowable fluctuation range into an interval mapping function to generate a reference parameter curve corresponding to the target dynamic response parameter.

[0063] Step S1233: constructing a linear compensation term and a nonlinear compensation term in the parameter adjustment function based on the reference parameter curve and the compensation weight coefficient, wherein the linear compensation term is used to perform proportional compensation on the steady-state deviation of the target dynamic response parameter, and the nonlinear compensation term is used to perform differential compensation on the transient fluctuation of the target dynamic response parameter.

[0064] Step S1234: superimpose the linear compensation term and the nonlinear compensation term to generate the parameter adjustment function, and use the parameter adjustment function to calculate the deviation of the real-time value of the target dynamic response parameter.

[0065] In this embodiment, during the movement of the quadruped robot, each optimization requirement in its optimization requirement set includes a requirement type identifier and a requirement constraint condition. For each optimization requirement, the target dynamic response parameter corresponding to the optimization requirement is first determined according to the requirement type identifier. Taking a complex scenario in which a quadruped robot crosses a series of different terrains (such as flat ground, slopes, potholes, etc.) and performs a rapid turning action at the same time as an example, when the requirement type identifier is an inertia delay optimization identifier, the corresponding target dynamic response parameter is the inertia delay parameter. In this scenario, the leg joints of the quadruped robot need to respond to control instructions quickly and accurately, and the size of the inertia delay directly affects the robot's movement flexibility and accuracy. If the inertia delay is too large, the leg joints cannot change the movement state in time after receiving the turning instruction, which may cause the robot to lose balance or fail to turn according to the predetermined trajectory. When the requirement type identifier is a torque fluctuation optimization identifier, the target dynamic response parameter is the torque fluctuation parameter. When the quadruped robot moves on different terrains, the load borne by the leg joints changes continuously, which causes torque fluctuations. For example, when climbing up a slope, a larger torque is needed to propel the robot forward, while when descending from the slope to flat ground, the torque requirement will decrease. If the torque fluctuation is too large, it will affect the robot's movement stability and may even cause damage to the joints and drive system. When the demand type is identified as the position deviation optimization identifier, the target dynamic response parameter is the position deviation parameter. When a quadruped robot is walking or performing a specific action, each leg joint has a predetermined position. If the position deviation is too large, it will affect the robot's overall posture and stability. For example, when crossing an obstacle, the leg joint needs to reach a specific position accurately to ensure that the robot passes smoothly, otherwise it may cause the robot to stumble.

[0066] Next, according to the parameter adjustment range defined in the demand constraint conditions, the constraint boundary of the target dynamic response parameter is extracted to obtain the allowable fluctuation range. The demand constraint conditions include time constraint conditions and space constraint conditions. For the inertia delay parameter, the maximum allowable delay time and the minimum allowable response time in the time constraint conditions are analyzed to determine the first time constraint interval. For example, when the quadruped robot performs a rapid turning action, according to the design specifications and motion performance requirements of the robot, the maximum allowable delay time is set to 0.05 seconds and the minimum allowable response time is 0.01 seconds. This determines that the first time constraint interval corresponding to the inertia delay parameter is [0.01, 0.05] seconds. The first time constraint interval limits the time fluctuation range of the inertia delay parameter. For the position deviation parameter, the maximum allowable position deviation distance in the space constraint conditions is analyzed to determine the first space constraint interval. Assuming that during the normal walking process of the quadruped robot, according to the structure and motion stability requirements of the robot, the maximum allowable position deviation distance is 3 mm, then the first space constraint interval corresponding to the position deviation parameter is [-3, 3] mm, and the first space constraint interval limits the spatial offset range of the position deviation parameter. For the torque fluctuation parameter, the minimum allowable torque fluctuation amplitude in the spatial constraint condition is also analyzed to determine the second spatial constraint interval. For example, considering the bearing capacity of the robot joint and drive system and the requirements for motion smoothness, the minimum allowable torque fluctuation amplitude is set to ±5%, then the second spatial constraint interval corresponding to the torque fluctuation parameter is [-5%, 5%], and the second spatial constraint interval limits the amplitude variation range of the torque fluctuation parameter. Based on these first time constraint intervals, first spatial constraint intervals, and second spatial constraint intervals, the first allowable fluctuation interval of the inertia delay parameter, the second allowable fluctuation interval of the position deviation parameter, and the third allowable fluctuation interval of the torque fluctuation parameter are constructed respectively.

[0067] The parameter adjustment function corresponding to the target dynamic response parameter is constructed based on the allowable fluctuation range and demand priority. Taking the task scenario of a quadruped robot performing complex terrain crossing and rapid action as an example, it is assumed that in this task, the demand priority of inertia delay optimization is the highest, torque fluctuation optimization is the second, and position deviation optimization is the lowest. The compensation weight coefficient corresponding to each allowable fluctuation range is determined according to the demand priority. Since the compensation weight coefficient is positively correlated with the demand priority, the compensation weight coefficient corresponding to the allowable fluctuation range of the inertia delay parameter is the largest, followed by the torque fluctuation parameter, and the position deviation parameter is the smallest. Taking the inertia delay parameter as an example, the upper and lower limits of its allowable fluctuation range are input into the interval mapping function to generate the benchmark parameter curve corresponding to the target dynamic response parameter. Assuming that the first allowable fluctuation range of the inertia delay parameter is [0.01, 0.05] seconds, a benchmark parameter curve reflecting the ideal change trend of the inertia delay parameter in the first allowable fluctuation range is obtained through the interval mapping function. Based on the benchmark parameter curve and the compensation weight coefficient, the linear compensation term and nonlinear compensation term in the parameter adjustment function are constructed. The linear compensation term is used to compensate the steady-state deviation of the target dynamic response parameter in proportion. For example, if the inertia delay parameter is continuously and stably higher than the middle value of the allowable fluctuation range, the linear compensation term will compensate the deviation according to the proportional relationship to make the inertia delay parameter approach the ideal value. The nonlinear compensation term is used to compensate the transient fluctuation of the target dynamic response parameter in differential way. When the inertia delay parameter changes suddenly (such as when the inertia suddenly increases during a quick turn, the delay suddenly increases), the nonlinear compensation term will compensate according to the rate of change, so that the fluctuation of the inertia delay parameter is suppressed. The linear compensation term and the nonlinear compensation term are superimposed to generate a parameter adjustment function, and the parameter adjustment function is used to calculate the deviation of the real-time value of the target dynamic response parameter. For example, in the actual movement of the quadruped robot, the inertia delay parameter measured in real time is 0.06 seconds. The deviation from the upper limit of the allowable fluctuation range of 0.05 seconds is calculated by the parameter adjustment function. The deviation includes the steady-state deviation and the deviation caused by transient fluctuation.

[0068] For the torque fluctuation parameter and position deviation parameter, the parameter adjustment function is constructed and the deviation is calculated in the same way as above. When the quadruped robot climbs up a slope, the torque fluctuation may exceed the allowable fluctuation range. The deviation is calculated by the corresponding parameter adjustment function, which reflects the degree of deviation between the actual torque fluctuation and the ideal fluctuation range. For the position deviation parameter, when the quadruped robot crosses an obstacle, if the position deviation of a leg joint exceeds the allowable fluctuation range, the deviation can also be accurately calculated by the parameter adjustment function.

[0069] After the deviation calculation is completed, the compensation signal calculation is performed, which includes the calculation of the reverse compensation for the parameter deviation value that exceeds the allowable fluctuation range. Taking the inertia delay parameter as an example, if the calculated deviation is positive (that is, the actual inertia delay is greater than the upper limit of the allowable fluctuation range), then the reverse compensation is a value that can reduce the inertia delay. The reverse compensation is calculated based on the parameter adjustment function and the actual deviation, and is intended to bring the inertia delay parameter back to the allowable fluctuation range. The same is true for the torque fluctuation parameter and the position deviation parameter. When the torque fluctuation exceeds the allowable fluctuation range, the corresponding reverse compensation is calculated to stabilize the torque fluctuation; when the position deviation exceeds the allowable fluctuation range, the reverse compensation is calculated to reduce the position deviation.

[0070] Finally, the reverse compensation amount is mapped to the corresponding control signal strength value, and the control signal strength values ​​are sorted according to the demand priority to generate a control signal sequence set. For the reverse compensation amount calculated by the inertia delay parameter, it is converted into a control signal strength value according to a certain mapping rule. The control signal strength value determines the size of the control signal when compensating for the inertia delay. Since the demand priority of inertia delay optimization is the highest, the control signal strength value will be in a more important position in the sorting. Similarly, the reverse compensation amount calculated for the torque fluctuation parameter and the position deviation parameter is also mapped to the control signal strength value and sorted according to their respective demand priorities. For example, in the control signal sequence set of the quadruped robot, the first is the control signal strength value corresponding to the inertia delay optimization, followed by the control signal strength value corresponding to the torque fluctuation optimization, and finally the control signal strength value corresponding to the position deviation optimization. The control signal sequence set generated in this way can accurately control the servo device of the quadruped robot according to the actual situation of the optimization requirements and the target dynamic response parameters, thereby improving the performance of the quadruped robot in complex motion scenes and ensuring the accuracy, stability and efficiency of its motion.

[0071] In a possible implementation, step S130 includes:

[0072] Step S131, parsing the execution timestamp and execution strength value corresponding to each control signal in the control signal sequence set.

[0073] Take the walking process of a quadruped robot as an example. When a quadruped robot takes a step, it involves the coordinated movement of multiple joints, and the movement of each joint is controlled by the corresponding control signal. For example, for the joint motor responsible for lifting the leg, the execution timestamp of its control signal may be set to 0.2 seconds after the robot starts moving. This timestamp accurately determines when the control signal begins to take effect on the joint. The execution strength value determines the strength and speed of the joint movement. If the execution strength value is high, it means that the joint will be lifted with greater force and faster speed, which is necessary for crossing higher obstacles or moving quickly on complex terrain.

[0074] Step S132: dividing the control signal sequence set into a plurality of control signal subsets according to the execution timestamp, wherein each control signal subset corresponds to a motion phase of the servo device.

[0075] A complete walking cycle of a quadruped robot can be divided into multiple motion phases, such as the leg-lifting preparation phase, leg-lifting phase, leg-stepping phase, and leg-releasing phase. For the leg-lifting preparation phase, the corresponding control signal subset contains signals that control the leg joints to prepare for movement from a static state. These signals have a specific execution timestamp range and execution intensity value range. In the leg-lifting phase, its control signal subset focuses on lifting the leg joints at an appropriate speed and strength. The control signals in this control signal subset are different from those in the leg-lifting preparation phase in terms of execution timestamps and execution intensity values. The control signal subset for each motion phase is carefully designed based on the motion logic and mechanical principles of the quadruped robot to ensure the coordinated movement of each joint at different stages.

[0076] Step S133, at the start time of each motion phase, the corresponding control signal subset is loaded into the signal buffer queue of the actuator, and the control signal in the signal buffer queue is power amplified according to the execution intensity value.

[0077] For example, when entering the leg-lifting phase, at the start of the leg-lifting phase, the relevant control signal subset is loaded into the signal buffer queue of the actuator. Assuming that in the control signal subset, there is a control signal whose execution strength value requires the leg joint to be lifted with greater force, in order to meet this requirement, the control signal in the signal buffer queue needs to be power amplified according to the execution strength value. This is like in a mechanical transmission system, in order to make a certain component move at a predetermined force and speed, sufficient energy input is required, and power amplification is to ensure that the control signal can drive the actuator to generate sufficient power.

[0078] Step S134: converting the amplified control signal into a physical drive instruction through the drive circuit of the actuator, and adjusting the motor speed, output torque and position feedback accuracy of the actuator based on the physical drive instruction.

[0079] In a possible implementation, step S134 includes:

[0080] Step S1341, obtaining the voltage amplitude, current frequency and pulse width modulation parameters in the amplified control signal.

[0081] Step S1342, calculating the target output voltage and target output current of the driving circuit according to the voltage amplitude and the current frequency.

[0082] Step S1343, generating a pulse waveform sequence matching the target output voltage and the target output current based on the pulse width modulation parameters.

[0083] Step S1344, inputting the pulse waveform sequence into the power switch device of the driving circuit to control the on time and off time of the power switch device.

[0084] Step S1345, converting the pulse waveform sequence into a physical drive instruction with a predetermined duty cycle and frequency through the alternating on and off operations of the power switch device.

[0085] For example, after the control signal in the leg-lifting stage is power-amplified, its voltage amplitude may reach 24 volts, the current frequency is 100 Hz, and the duty cycle in the pulse width modulation parameters is 60%. The target output voltage and target output current of the drive circuit are calculated based on the voltage amplitude and current frequency. In this example, the target output voltage and target output current values ​​suitable for the drive circuit are calculated based on the voltage amplitude of 24 volts and the current frequency of 100 Hz through a specific circuit calculation formula. These values ​​are key parameters to ensure the normal operation of the motor. Then, a pulse waveform sequence matching the target output voltage and target output current is generated based on the pulse width modulation parameters. The pulse waveform sequence contains various information for controlling the movement of the motor. The pulse waveform sequence is input into the power switch device of the drive circuit to control the on time and off time of the power switch device. For example, when a pulse in the pulse waveform sequence arrives, the power switch device is turned on and the current passes through the motor. When the pulse disappears, the power switch device is turned off. Through the alternating on and off operation of the power switch device, the pulse waveform sequence is converted into a physical drive instruction with a predetermined duty cycle and frequency. The physical drive instruction directly acts on the motor. Depending on the duty cycle and frequency, the motor speed, output torque and position feedback accuracy will change accordingly. For example, a higher duty cycle may increase the motor speed, thereby lifting the leg joint faster, while also affecting the output torque and position feedback accuracy, ensuring that the leg joint can accurately reach the predetermined position and provide sufficient support.

[0086] Step S135 , monitoring the real-time motion state of the actuator in the motion phase, and when detecting an abnormal motion state, interrupting the execution of the current control signal subset and triggering an abnormal recovery mechanism.

[0087] For example, during the leg-lifting phase of the quadruped robot, if the motor speed of the leg joint is detected to suddenly drop below the normal range of motion speed, this may be due to motor failure, sudden increase in load (such as the leg encountering an unexpected obstacle) or other reasons. Once this abnormal motion state is detected, the execution of the control signal subset of the current leg-lifting phase will be immediately interrupted. For example, if there are subsequent control signals in the control signal subset that have not yet been executed, these control signals will be suspended to prevent further erroneous operations. At the same time, the abnormal recovery mechanism is triggered, which may include re-evaluating the current motion state, adjusting the control signal parameters, reducing the motion speed or taking other measures to ensure the stability of the quadruped robot. If the abnormality is caused by a sudden increase in load, the execution intensity value of the control signal may be adjusted to increase the output torque to overcome the additional load so that the leg joint can continue to move normally. If it is a motor failure, it may switch to a spare motor or start a fault diagnosis program to restore the normal operation of the quadruped robot as soon as possible.

[0088] The process of converting the amplified control signal into a physical drive instruction through the drive circuit of the actuator is further explained by taking the leg joint movement of the quadruped robot during walking as an example. When the amplified control signal enters the drive circuit, it is crucial to obtain the voltage amplitude, current frequency and pulse width modulation parameters. Assume that in the leg-stepping stage of the quadruped robot, the voltage amplitude of the amplified control signal is 36 volts, the current frequency is 80 Hz, and the duty cycle in the pulse width modulation parameter is 50%. First, the target output voltage and target output current of the drive circuit are calculated based on the voltage amplitude and current frequency. Through the electrical calculation formula, according to the voltage amplitude of 36 volts and the current frequency of 80 Hz, the target voltage and target current values ​​that the drive circuit should output are determined. These calculated target values ​​are to match the working characteristics of the motor and ensure that the motor can operate under appropriate electrical conditions.

[0089] A pulse waveform sequence matching the target output voltage and the target output current is generated based on the pulse width modulation parameters. The generation of the pulse waveform sequence is based on the principle of pulse width modulation technology. A specific pulse sequence is constructed according to parameters such as duty cycle and frequency. Each pulse in the pulse sequence contains information for controlling the motor. For example, the width of the pulse determines the on-time of the power switch device, thereby affecting the input energy and operating state of the motor. Then, the pulse waveform sequence is input to the power switch device of the drive circuit to control the on-time and off-time of the power switch device. For example, when a high-level pulse in the pulse waveform sequence arrives, the power switch device is turned on, the current flows from the power supply to the motor, and the motor starts to rotate. When the pulse becomes a low level, the power switch device is turned off, the current stops flowing, and the motor stops rotating or starts to decelerate. Through the alternating on and off operations of the power switch device, the pulse waveform sequence is converted into a physical drive instruction with a predetermined duty cycle and frequency. The physical drive instruction directly acts on the motor, so that the motor adjusts its own speed, output torque and position feedback accuracy according to the instruction. For example, if the duty cycle in the physical drive instruction increases, the average input voltage of the motor will increase, resulting in faster motor speed and a corresponding increase in output torque. The position feedback accuracy also needs to be adjusted accordingly to ensure that the actual position of the leg joints matches the expected position and that the quadruped robot can walk stably. In this process, any problem in any link may cause abnormal movement of the quadruped robot, so precise control and monitoring are the key to ensuring the normal operation of the quadruped robot.

[0090] In a possible implementation, step S141 includes:

[0091] Step S1411: collecting the acceleration feedback signal and the angular velocity feedback signal of the actuator through the inertial measurement unit in the servo device.

[0092] During the movement of the quadruped robot, various sensors in the servo device ensure that the feedback signals of the actuator (i.e., the joint motors of the quadruped robot and its related transmission components, etc.) during the dynamic response process can be accurately collected. First, the acceleration feedback signal and angular velocity feedback signal of the actuator are collected through the inertial measurement unit in the servo device. During the walking process of the quadruped robot, when the leg joints start to move, whether from static to lifting, moving or putting down, changes in acceleration and angular velocity will occur. The inertial measurement unit can accurately sense these changes and generate corresponding feedback signals. For example, when the legs of the quadruped robot start to accelerate and lift from the static state when standing to prepare to step over a small obstacle, the inertial measurement unit will detect the acceleration value and angular velocity value during the acceleration process. The acceleration feedback signal reflects the magnitude and direction of the leg joint acceleration, and the angular velocity feedback signal reflects the change in the speed of joint rotation.

[0093] Step S1412: collecting the real-time torque feedback signal of the actuator through the torque sensor in the servo device.

[0094] In each stage of the quadruped robot's movement, the torque on the leg joints is constantly changing. For example, when the quadruped robot walks up a slope, in order to overcome gravity and push the robot forward, the leg joints need a larger torque. The torque sensor can measure the magnitude of this torque in real time and generate a feedback signal. This real-time torque feedback signal is crucial for evaluating the load condition of the joints and the power distribution of the entire robot.

[0095] Step S1413, collecting the actual position feedback signal of the actuator through the position encoder in the servo device.

[0096] When a quadruped robot moves, each leg joint has its predetermined motion trajectory and target position. The position encoder can accurately measure the actual position of the joint and generate feedback signals. For example, during a quadruped robot's complete walking cycle, the position encoder continuously monitors whether the leg joint has accurately reached the predetermined lifting height, stepping distance, and lowering position.

[0097] Step S1414, performing noise filtering processing on the acceleration feedback signal, angular velocity feedback signal, real-time torque feedback signal and actual position feedback signal to obtain denoised acceleration data, angular velocity data, torque data and position data.

[0098] However, after collecting these acceleration feedback signals, angular velocity feedback signals, real-time torque feedback signals and actual position feedback signals, due to various interference factors in the actual environment, these signals inevitably contain noise components, which need to be filtered to obtain accurate signal data.

[0099] Wherein, step S1414 includes:

[0100] Step S1414-1, identifying the high-frequency vibration noise component in the acceleration feedback signal, and attenuating the high-frequency vibration noise component using a low-pass filter.

[0101] During the movement of the quadruped robot, high-frequency vibration noise may be mixed into the acceleration feedback signal due to friction between mechanical parts, electromagnetic vibration of the motor, and other reasons. For example, when the motor of the leg joint is running, the electromagnetic force inside the motor may cause some tiny high-frequency vibrations, which will be reflected in the acceleration feedback signal. The low-pass filter sets a suitable cutoff frequency and only allows signal components below the cutoff frequency to pass, thereby effectively attenuating the high-frequency vibration noise components. For example, assuming that the useful signal frequency in the acceleration feedback signal is mainly concentrated in the range of 0-100Hz, and the frequency of the high-frequency vibration noise component is above 500Hz, setting the cutoff frequency of the low-pass filter to 150Hz can attenuate most of the high-frequency vibration noise components and obtain a relatively pure acceleration feedback signal.

[0102] Step S1414-2, detecting the instantaneous pulse interference component in the angular velocity feedback signal, and smoothing the instantaneous pulse interference component using a sliding window mean filtering algorithm.

[0103] During the movement of the quadruped robot, some unexpected situations may occur, causing instantaneous pulse interference components in the angular velocity feedback signal. For example, when the leg joint is suddenly subjected to a short external force impact during the movement, such as accidentally hitting an object next to it, the angular velocity feedback signal will instantly generate a large pulse. The sliding window mean filter algorithm sets a suitable window size and calculates the average value of the data in the sliding window to replace the data in the center of the window. Assuming that the size of the sliding window is 5 data points, when instantaneous pulse interference is detected, the average value of other normal data points in the window will be used to replace the data points containing pulse interference, thereby achieving smoothing of the instantaneous pulse interference components and obtaining a more stable angular velocity feedback signal.

[0104] Step S1414-3, extracting the periodic fluctuation noise in the real-time torque feedback signal, and applying an adaptive notch filter to eliminate the periodic fluctuation noise.

[0105] During the movement of the quadruped robot, the real-time torque feedback signal may contain periodic fluctuation noise due to some inherent characteristics of the motor or periodic interference of the external environment. For example, the electromagnetic torque of the motor may have periodic fluctuations of a certain frequency, or when the quadruped robot walks on regular terrain (such as periodic bumps or depressions), the periodic changes of the external load will also cause periodic fluctuations in the torque feedback signal. The adaptive notch filter can automatically detect and adapt to the frequency of this periodic fluctuation noise, and then eliminate it. Assuming that the frequency of the periodic fluctuation noise is 100Hz, the adaptive notch filter can accurately locate the frequency component, and by adjusting the parameters of the filter, the signal component with a frequency of 100Hz can be removed from the real-time torque feedback signal to obtain a pure torque feedback signal.

[0106] Step S1414-4, analyzing the quantization error component in the actual position feedback signal, and compensating and correcting the quantization error component through an interpolation algorithm.

[0107] During the position measurement process of the quadruped robot, the position encoder will produce quantization errors due to its own limited resolution. For example, the position encoder may discretize the actual continuous position change into a certain quantization unit. When the actual position change is between two quantization units, a quantization error will be generated. The interpolation algorithm estimates and compensates for the quantization error component based on a certain mathematical model by using known position data points. For example, if the quantization unit of the position encoder is 1 mm, when the actual position is at 1.2 mm, but the position encoder can only feedback 1 mm or 2 mm, the interpolation algorithm can compensate for the quantization error based on the relationship between the surrounding position data points to obtain a more accurate actual position feedback signal.

[0108] Step S1414-5, using the filtered, smoothed, eliminated and corrected signal data as denoised acceleration data, angular velocity data, torque data and position data.

[0109] Step S1415, converting the denoised acceleration data into actual inertia delay parameters, combining the denoised angular velocity data with the position data to generate actual position deviation parameters, and converting the denoised torque data into actual torque fluctuation parameters.

[0110] In the present embodiment, the acceleration data after denoising is converted into actual inertia delay parameters, and the conversion process is based on the kinematic and dynamic models of the quadruped robot. For example, according to the mass, moment of inertia and actual acceleration data of the leg joints, the actual inertia delay parameters are obtained by a specific calculation formula. The angular velocity data after denoising is combined with the position data to generate the actual position deviation parameter, because the change of angular velocity will affect the motion trajectory of the leg joints, and the deviation of the joint relative to the predetermined position can be accurately calculated in combination with the actual position data. For example, the theoretical position that the joint should reach within a certain period of time is calculated according to the angular velocity data, and compared with the actual position data to obtain the position deviation. The torque data after denoising is converted into the actual torque fluctuation parameter, which reflects the fluctuation of the actual torque relative to the ideal stable torque, which is of great significance for evaluating the load stability and dynamic performance of the joint. Through these accurately collected and processed feedback signal parameters, the actual dynamic performance of the quadruped robot actuator can be accurately evaluated.

[0111] In a possible implementation, step S142 includes:

[0112] Step S1421, calculating the absolute value of the difference between the inertia delay deviation value, torque fluctuation deviation value and position deviation value in the current dynamic performance index and the corresponding inertia delay threshold value, torque fluctuation threshold value and position deviation threshold value in the preset performance threshold value.

[0113] In this embodiment, during the motion of the quadruped robot, in order to ensure its precise motion control and good performance, it is necessary to dynamically adjust the weight of each control signal in the control signal sequence set according to the degree of difference between the actual dynamic performance index and the preset performance threshold. First, the absolute value of the difference between the inertia delay deviation value, torque fluctuation deviation value and position deviation value in the current dynamic performance index and the corresponding inertia delay threshold, torque fluctuation threshold and position deviation threshold in the preset performance threshold is calculated. For example, when the quadruped robot performs tasks of crossing complex terrain (including slopes of different slopes, potholes of different sizes, etc.) and complex actions (such as fast turning, jumping, etc.), the preset inertia delay threshold is set to 0.03 seconds, and the actual inertia delay value in the current dynamic performance index is 0.05 seconds, then the absolute value of the difference between the inertia delay deviation value and the preset inertia delay threshold is 0.02 seconds. For torque fluctuation, assuming that the preset torque fluctuation threshold is ±10%, and the actual measured torque fluctuation value is ±15%, the absolute value of the difference between the torque fluctuation deviation value and the preset torque fluctuation threshold is 5%. For the position deviation, if the preset position deviation threshold is 2 mm and the actually measured position deviation value is 3 mm, then the absolute value of the difference between the position deviation value and the preset position deviation threshold is 1 mm.

[0114] Step S1422: determining a weight adjustment coefficient corresponding to each control signal according to the absolute value of the difference, wherein the weight adjustment coefficient is positively correlated with the absolute value of the difference.

[0115] Since the weight adjustment coefficient is positively correlated with the absolute value of the difference, that is, the larger the absolute value of the difference is, the larger the weight adjustment coefficient is. Taking inertia delay as an example, if the absolute value of the difference between the inertia delay deviation value and the preset threshold is 0.02 seconds, this indicates that the actual inertia delay exceeds the preset range by a large amount, then the corresponding weight adjustment coefficient will be larger, which means that the control signal related to the inertia delay needs to be adjusted to a greater extent. The same is true for torque fluctuation and position deviation. If the absolute value of the difference between the torque fluctuation deviation value and the preset threshold is large, or the absolute value of the difference between the position deviation value and the preset threshold is large, then their corresponding weight adjustment coefficients will also increase accordingly.

[0116] Step S1423: scaling the strength value of each control signal in the control signal sequence set based on the weight adjustment coefficient to generate an intermediate control signal sequence.

[0117] In the control signal system of the quadruped robot, each control signal has its specific intensity value, which determines the motion characteristics of the actuator (such as the leg joint motor), such as speed, force, etc. For example, for the control signal that controls the lifting speed of the leg joint, its original intensity value may be set to lift the joint at a certain angular velocity. If the weight adjustment coefficient related to the inertia delay is large, the intensity value of the control signal is scaled according to the weight adjustment coefficient, which may increase the intensity value to speed up the lifting speed of the leg joint, thereby reducing the inertia delay. Similarly, for the control signals related to torque fluctuations and position deviations, their intensity values ​​are scaled according to their respective weight adjustment coefficients to generate an intermediate control signal sequence, in which each control signal intensity value in the intermediate control signal sequence has been adjusted to adapt to the difference between the actual dynamic performance and the preset performance threshold, aiming to optimize the motion performance of the quadruped robot.

[0118] Step S1424: offset-compensate the execution timestamp of each control signal in the intermediate control signal sequence so that the execution time of the scaled control signal matches the dynamic response phase of the servo device.

[0119] In a possible implementation, step S1424 includes:

[0120] Step S1424-1, obtaining the phase detection signal of the servo device in the current motion cycle, and analyzing the zero-crossing point and the peak point in the phase detection signal.

[0121] In the leg joint motion cycle of the quadruped robot, the phase detection signal reflects the periodic changes of the joint motion. For example, when the leg joint starts to lift up from the lowest position, the phase detection signal will start a new cycle, and there will be specific zero crossing points and peak points in this cycle. The zero crossing point may correspond to the change of the joint motion direction or the conversion of the force balance state, and the peak point may correspond to the maximum point of the joint motion speed or the maximum point of the torque demand, etc.

[0122] Step S1424-2, determining the dynamic response phase angle of the servo device according to the zero-crossing point and the peak point, and calculating the theoretical optimal execution time of each control signal based on the dynamic response phase angle.

[0123] For example, for the control signal that controls the fine-tuning of the leg joints during the ascent, based on the motion cycle, speed curve and current phase angle of the leg joints, it is calculated that the theoretical optimal execution time of the control signal should be when the joint rises to a certain height and the speed begins to slow down. This theoretical optimal execution time can most effectively fine-tune the joint position while avoiding negative impacts on the stability of the joint movement.

[0124] Step S1424-3: compare the time offset between the original execution timestamp of each control signal in the intermediate control signal sequence and the theoretical optimal execution time.

[0125] Step S1424-4: if the time offset is greater than a preset offset threshold, the original execution timestamp is corrected according to the theoretical optimal execution time.

[0126] Step S1424-5: If the time offset is less than or equal to the preset offset threshold, the original execution timestamp is kept unchanged.

[0127] For example, the original execution timestamp of a control signal is set to be executed at 0.3 seconds after the leg joint starts to lift, but the theoretical optimal execution time calculated is 0.25 seconds, so the time offset is 0.05 seconds. If the time offset is greater than the preset offset threshold (assuming the preset offset threshold is 0.03 seconds), the original execution timestamp is corrected according to the theoretical optimal execution time. This is done to ensure that the control signal is executed at the most appropriate time to optimize the movement effect of the quadruped robot. If the time offset is less than or equal to the preset offset threshold, for example, the time offset is 0.02 seconds, then the original execution timestamp is kept unchanged, because in this case, the deviation between the original execution timestamp and the theoretical optimal execution time is small, and will not have a significant adverse effect on the movement of the quadruped robot.

[0128] Step S1425: taking the control signal sequence after intensity scaling and timestamp shifting as the optimized control signal sequence set.

[0129] In this embodiment, the optimized control signal sequence set comprehensively considers the difference between the current dynamic performance index and the preset performance threshold, and can more accurately control the servo device of the quadruped robot through dynamic weight adjustment (including proportional scaling of the intensity value and offset compensation of the execution timestamp), thereby improving its performance in complex motion scenarios. For example, when the quadruped robot performs the task of crossing complex terrain and complex movements again, the optimized control signal sequence set can make the movement of the leg joints more coordinated and precise, reduce inertial delay, torque fluctuation and position deviation, and other problems, thereby improving the overall motion stability, speed and accuracy of the quadruped robot and other performance indicators.

[0130] Figure 2 The schematic diagram shows exemplary hardware and software components of the algorithm-controlled servo system optimization system 100 provided by some embodiments of the present application that can implement the concept of the present application. For example, the processor 120 can be used in the algorithm-controlled servo system optimization system 100 and used to perform the functions in the present application.

[0131] The algorithm-controlled servo system optimization system 100 can be a general-purpose server or a special-purpose server, both of which can be used to implement the algorithm-controlled servo system optimization method of the present application. Although only one server is shown in the present application, for convenience, the functions described in the present application can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0132] For example, the algorithm-controlled servo system optimization system 100 may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and storage media 140 in different forms, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the algorithm-controlled servo system optimization system 100 may also include program instructions stored in ROM, RAM, or other types of non-temporary storage media, or any combination thereof. The method of the present application can be implemented according to these program instructions. The algorithm-controlled servo system optimization system 100 also includes an input / output (I / O) interface 150 between the computer and other input / output devices.

[0133] For ease of explanation, only one processor is described in the algorithm-controlled servo system optimization system 100. However, it should be noted that the algorithm-controlled servo system optimization system 100 in the present application may also include multiple processors, so the steps performed by one processor described in the present application may also be performed jointly or individually by multiple processors. For example, if the processor of the algorithm-controlled servo system optimization system 100 executes step A and step B, it should be understood that step A and step B may also be performed jointly by two different processors or individually in one processor. For example, the first processor executes step A, the second processor executes step B, or the first processor and the second processor execute steps A and B together.

[0134] In addition, an embodiment of the present invention further provides a readable storage medium, in which computer executable instructions are preset. When a processor executes the computer executable instructions, the servo system optimization method based on algorithm control as described above is implemented.

[0135] It should be noted that in order to simplify the description of the present invention and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, various features are sometimes combined into one embodiment, drawing or description thereof.

Claims

1. A servo system optimization method based on algorithm control, characterized in that: The method comprises: Acquire a real-time dynamic response parameter set generated by the servo device during dynamic operation, and an optimization requirement set corresponding to the servo device; wherein the real-time dynamic response parameter set includes an inertia delay parameter, a torque fluctuation parameter, and a position deviation parameter generated by the servo device when executing a target motion trajectory, and the optimization requirement set includes a first optimization requirement for the inertia delay parameter, a second optimization requirement for the torque fluctuation parameter, and a third optimization requirement for the position deviation parameter; According to the demand priority corresponding to each optimization demand in the optimization demand set, multi-level control signal generation is performed on the real-time dynamic response parameter set to obtain a control signal sequence set matching each optimization demand; wherein the multi-level control signal generation includes generating a first compensation signal for the inertia delay parameter, generating a second compensation signal for the torque fluctuation parameter, and generating a third compensation signal for the position deviation parameter; Inputting the control signal sequence set to the actuator of the servo device to drive the actuator to dynamically respond according to the updated motion trajectory; The control signal sequence set is optimized based on the feedback signal set generated by the actuator during the dynamic response process.

2. The servo system optimization method based on algorithm control according to claim 1, characterized in that: The optimizing operation on the control signal sequence set based on the feedback signal set generated by the actuator in the dynamic response process includes: Collecting a feedback signal set generated by the actuator during the dynamic response process, and determining a current dynamic performance index of the servo device based on an actual inertia delay parameter, an actual torque fluctuation parameter, and an actual position deviation parameter in the feedback signal set; According to the difference between the current dynamic performance index and the preset performance threshold, each control signal in the control signal sequence set is dynamically weighted to generate an optimized control signal sequence set, and the optimized control signal sequence set is iteratively input to the actuator.

3. The servo system optimization method based on algorithm control according to claim 1, characterized in that: Each optimization requirement in the optimization requirement set includes a requirement type identifier and a requirement constraint condition. The multi-level control signal generation is performed on the real-time dynamic response parameter set according to the requirement priority corresponding to each optimization requirement in the optimization requirement set to obtain a control signal sequence set matching each optimization requirement, including: For each optimization requirement in the optimization requirement set, determining a target dynamic response parameter corresponding to the optimization requirement according to the requirement type identifier; wherein the requirement type identifier includes an inertia delay optimization identifier, a torque fluctuation optimization identifier, and a position deviation optimization identifier; According to the parameter adjustment range defined in the demand constraint condition, the constraint boundary of the target dynamic response parameter is extracted to obtain the allowable fluctuation range of the target dynamic response parameter; Based on the allowable fluctuation range and the demand priority, construct a parameter adjustment function corresponding to the target dynamic response parameter, and call the parameter adjustment function to perform compensation signal calculation on the target dynamic response parameter; wherein the compensation signal calculation includes calculating a reverse compensation amount for a parameter deviation value that exceeds the allowable fluctuation range; The reverse compensation amount is mapped to a corresponding control signal strength value, and the control signal strength values ​​are sorted according to the demand priority to generate the control signal sequence set.

4. The servo system optimization method based on algorithm control according to claim 3 is characterized in that: The demand constraint condition includes a time constraint condition and a space constraint condition. The step of extracting the constraint boundary of the target dynamic response parameter according to the parameter adjustment range defined in the demand constraint condition to obtain the allowable fluctuation range of the target dynamic response parameter includes: Analyze the maximum allowable delay time and the minimum allowable response time in the time constraint condition, and determine a first time constraint interval corresponding to the inertia delay parameter; Analyze the maximum allowable position deviation distance and the minimum allowable torque fluctuation amplitude in the spatial constraint condition, and determine a first spatial constraint interval corresponding to the position deviation parameter and a second spatial constraint interval corresponding to the torque fluctuation parameter; According to the first time constraint interval, the first space constraint interval and the second space constraint interval, a first allowable fluctuation interval of the inertia delay parameter, a second allowable fluctuation interval of the position deviation parameter and a third allowable fluctuation interval of the torque fluctuation parameter are respectively constructed; wherein the first allowable fluctuation interval is used to limit the time fluctuation range of the inertia delay parameter, the second allowable fluctuation interval is used to limit the spatial offset range of the position deviation parameter, and the third allowable fluctuation interval is used to limit the amplitude variation range of the torque fluctuation parameter; And, constructing a parameter adjustment function corresponding to the target dynamic response parameter based on the allowable fluctuation range and the demand priority includes: Determine a compensation weight coefficient corresponding to each allowable fluctuation interval according to the demand priority, wherein the compensation weight coefficient is positively correlated with the demand priority; Inputting the upper limit value and the lower limit value of the allowable fluctuation range into an interval mapping function to generate a reference parameter curve corresponding to the target dynamic response parameter; Based on the reference parameter curve and the compensation weight coefficient, a linear compensation term and a nonlinear compensation term in the parameter adjustment function are constructed; wherein the linear compensation term is used to perform proportional compensation on the steady-state deviation of the target dynamic response parameter, and the nonlinear compensation term is used to perform differential compensation on the transient fluctuation of the target dynamic response parameter; The linear compensation term and the nonlinear compensation term are superimposed to generate the parameter adjustment function, and the parameter adjustment function is used to calculate the deviation of the real-time value of the target dynamic response parameter.

5. The servo system optimization method based on algorithm control according to claim 1, characterized in that: The step of inputting the control signal sequence set to the actuator of the servo device to drive the actuator to dynamically respond according to the updated motion trajectory includes: Parsing the execution timestamp and execution strength value corresponding to each control signal in the control signal sequence set; Dividing the control signal sequence set into a plurality of control signal subsets according to the execution timestamp, wherein each control signal subset corresponds to a motion phase of the servo device; At the start of each motion phase, a corresponding subset of control signals is loaded into a signal buffer queue of the actuator, and power amplification of the control signals in the signal buffer queue is performed according to the execution intensity value; The amplified control signal is converted into a physical drive instruction by the drive circuit of the actuator, and the motor speed, output torque and position feedback accuracy of the actuator are adjusted based on the physical drive instruction; The real-time motion state of the actuator in the motion phase is monitored, and when an abnormal motion state is detected, the execution of the current control signal subset is interrupted and an abnormal recovery mechanism is triggered.

6. The servo system optimization method based on algorithm control according to claim 5, characterized in that: The step of converting the amplified control signal into a physical drive instruction through the drive circuit of the actuator includes: Obtaining voltage amplitude, current frequency and pulse width modulation parameters in the amplified control signal; Calculating a target output voltage and a target output current of the drive circuit according to the voltage amplitude and the current frequency; generating a pulse waveform sequence matching the target output voltage and the target output current based on the pulse width modulation parameters; Inputting the pulse waveform sequence into the power switch device of the drive circuit to control the on time and off time of the power switch device; The pulse waveform sequence is converted into a physical driving instruction with a predetermined duty cycle and frequency through the alternating on and off operations of the power switch device.

7. The servo system optimization method based on algorithm control according to claim 2, characterized in that: The collecting of the feedback signal set generated by the actuator during the dynamic response process includes: Collecting the acceleration feedback signal and angular velocity feedback signal of the actuator through the inertial measurement unit in the servo device; Collecting a real-time torque feedback signal of the actuator through a torque sensor in the servo device; Acquiring an actual position feedback signal of the actuator through a position encoder in the servo device; Performing noise filtering on the acceleration feedback signal, angular velocity feedback signal, real-time torque feedback signal and actual position feedback signal to obtain denoised acceleration data, angular velocity data, torque data and position data; converting the denoised acceleration data into actual inertial delay parameters, combining the denoised angular velocity data with the position data to generate actual position deviation parameters, and converting the denoised torque data into actual torque fluctuation parameters; The noise filtering process of the acceleration feedback signal, the angular velocity feedback signal, the real-time torque feedback signal and the actual position feedback signal includes: Identifying high-frequency vibration noise components in the acceleration feedback signal, and attenuating the high-frequency vibration noise components using a low-pass filter; Detecting the instantaneous pulse interference component in the angular velocity feedback signal, and smoothing the instantaneous pulse interference component using a sliding window mean filtering algorithm; Extracting periodic fluctuation noise from the real-time torque feedback signal, and applying an adaptive notch filter to eliminate the periodic fluctuation noise; Analyzing the quantization error component in the actual position feedback signal, and compensating and correcting the quantization error component through an interpolation algorithm; The signal data after filtering, smoothing, elimination and correction are used as denoised acceleration data, angular velocity data, torque data and position data.

8. The servo system optimization method based on algorithm control according to claim 2, characterized in that: The dynamically adjusting the weight of each control signal in the control signal sequence set according to the difference between the current dynamic performance indicator and the preset performance threshold comprises: Calculating the absolute value of the difference between the inertia delay deviation value, the torque fluctuation deviation value and the position deviation value in the current dynamic performance index and the inertia delay threshold value, the torque fluctuation threshold value and the position deviation threshold value corresponding to the preset performance threshold value; Determining a weight adjustment coefficient corresponding to each control signal according to the absolute value of the difference, wherein the weight adjustment coefficient is positively correlated with the absolute value of the difference; Scaling the strength value of each control signal in the control signal sequence set based on the weight adjustment coefficient to generate an intermediate control signal sequence; Performing offset compensation on the execution timestamp of each control signal in the intermediate control signal sequence so that the execution time of the scaled control signal matches the dynamic response phase of the servo device; The control signal sequence after intensity scaling and time stamp shifting is used as the optimized control signal sequence set.

9. The servo system optimization method based on algorithm control according to claim 8, characterized in that: The offset compensation of the execution timestamp of each control signal in the intermediate control signal sequence includes: Acquire a phase detection signal of the servo device in a current motion cycle, and analyze a zero-crossing point and a peak point in the phase detection signal; Determine the dynamic response phase angle of the servo device according to the zero-crossing point and the peak point, and calculate the theoretical optimal execution time of each control signal based on the dynamic response phase angle; Comparing the time offset between the original execution timestamp of each control signal in the intermediate control signal sequence and the theoretical optimal execution time; If the time offset is greater than a preset offset threshold, the original execution timestamp is corrected according to the theoretical optimal execution time; If the time offset is less than or equal to the preset offset threshold, the original execution timestamp is kept unchanged.

10. A servo system optimization system based on algorithm control, characterized in that: The algorithm-controlled servo system optimization system includes a processor and a memory, wherein the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the algorithm-controlled servo system optimization method described in any one of claims 1 to 9.

Images