Working condition self-adaptive joint module motor control method for quadruped robot

Abstract

The invention discloses a working condition self-adaptive joint module motor control method for a quadruped robot, which comprises the following steps: S1, after a system is started, loading a preset standard parameter library of six working conditions of constant-speed walking on the flat ground, acceleration / deceleration on the flat ground, climbing, downhill, running and jumping and obstacle crossing, and calibrating a zero point of a multi-source sensor; a flat ground constant-speed walking working condition control mode is entered by default; and S2, collecting motor parameter data in real time through a multi-source working condition sensing module, carrying out fusion analysis on the collected motor parameter data by adopting a combinatorial algorithm, and carrying out working condition identification once every 5-10ms. Core parameters are collected through fusion of the multi-source sensor, a feature extraction and fuzzy neural network combined algorithm is combined, working condition recognition is completed every 5-10 ms, the recognition accuracy of six kinds of working conditions is high, the problem of working condition distinguishing fuzziness caused by dependence of a single parameter in the prior art is solved, and the working condition recognition accuracy is improved. And differential control requirements of complex scenes such as flat ground, climbing, running and jumping and the like can be accurately met.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of robots, and particularly relates to a working condition adaptive joint module motor control method for a quadruped robot. BACKGROUND

[0002] At present, as a complex dynamic system, the joint motor control performance of a quadruped robot directly determines the motion stability, adaptability and endurance. The prior art has three core defects, which are difficult to meet the application requirements in complex scenes: firstly, the working condition recognition accuracy is insufficient: in the prior art, the joint motor control only relies on the speed signal for fuzzy control, lacks the logic of “multi-source data fusion and working condition classification”, and cannot accurately distinguish the subdivided working conditions such as “flat ground uniform speed / acceleration, climbing / descending, running / jumping and obstacle crossing”, resulting in serious deviation of control parameter adaptation. Secondly, for the mechanical switching type technology, these joint motors are controlled by gear set switching speed ratio, which belongs to hardware structure adjustment, and not only has a response delay of greater than or equal to 100 ms, but also has problems such as mechanical wear, and cannot dynamically adapt the core parameters of motor control (such as PID, torque threshold, etc.); for the multi-sensor fusion type technology, it focuses on gait planning and does not establish a direct correlation between perception data and motor control parameters, and the control accuracy is limited by the gait planning algorithm, rather than the optimization of the motor execution layer. Thirdly, for the existing parameter adjustment type technology, it only optimizes the control accuracy or energy consumption, and cannot meet the differentiated needs of complex working conditions, such as low energy consumption for flat ground walking, high dynamic response for running and jumping, and strong load capacity for climbing, so single parameter control will lead to insufficient response in energy saving, high energy consumption in high response, and lack of vibration suppression and mechanical characteristic compensation in high motion intensity (such as running, jumping and obstacle crossing), so the joint motor is prone to resonance, control error increases, and even burns out due to overload.

[0003] In summary, the prior art does not form a complete control logic of “multi-source working condition perception-parameter adaptive mapping-multi-dimensional dynamic compensation-mechanical and electrical cooperation”, and cannot solve the problems of accurate control, multi-objective balance and high reliability of the quadruped robot in all working conditions, and there is a clear technical gap. SUMMARY

[0004] The purpose of the present application is to overcome the technical defects that the current quadruped robot cannot realize accurate control, multi-objective balance and high reliability in all working conditions, and provide a working condition adaptive joint module motor control method for a quadruped robot.

[0005] The purpose of the present application is achieved by the following technical scheme: a working condition adaptive joint module motor control method for a quadruped robot, comprising the following steps:

[0006] S1, after the system is started, a preset flat ground uniform walking, flat ground acceleration / deceleration, climbing, descending, running, jumping and obstacle crossing six kinds of working condition benchmark parameter library are loaded, the zero points of the multi-source sensors are calibrated, and the flat ground uniform walking working condition control mode is entered by default;

[0007] S2, the motor parameter data is collected in real time through the multi-source working condition sensing module, the combination algorithm is used to analyze the collected motor parameter data, the working condition is identified every 5-10ms, and the identified working condition is divided into one of the above six kinds of working conditions;

[0008] S3, if the system identifies the working condition switching, the corresponding benchmark parameter is called, and the benchmark parameter is dynamically corrected based on the real-time collected motor parameter;

[0009] S4, the load disturbance compensation, vibration suppression compensation and mechanical characteristic compensation instructions are output through the dynamic compensation control module and superimposed into the motor control signal;

[0010] S5, based on the joint angle and motor current feedback data, the control parameters are continuously adjusted to ensure that the joint motion angle error is less than or equal to 0.5°;

[0011] S6, the motor current, joint torque and motor temperature are monitored in real time, when overload, overtemperature or sensor failure is triggered, the protection logic is executed and the warning signal is sent.

[0012] Further, the "multi-source sensor" in step S1 comprises at least joint angle sensor, torque sensor, motor current sensor, IMU attitude sensor and six-dimensional force sensor.

[0013] The "motor parameter data" in step S2 comprises at least joint angle / angle velocity, output torque, motor stator current, robot body attitude angle and ground contact pressure.

[0014] The "benchmark parameter" in step S3 comprises at least fuzzy adaptive PID parameter group, torque upper threshold, speed response bandwidth and current protection threshold; the "real-time collected motor parameter" comprises at least load torque, motor temperature and joint vibration acceleration data.

[0015] The "combination algorithm" in step S2 is a feature extraction algorithm and a fuzzy neural network algorithm.

[0016] In order to better implement the present application, in the multi-source working condition sensing module, the accuracy of the joint angle sensor is less than or equal to 0.1°, the sampling rate of the IMU attitude sensor is greater than or equal to 1000Hz, and the resolution of the six-dimensional force sensor is less than or equal to 0.1N.

[0017] The fuzzy neural network algorithm is provided with an input layer, a hidden layer and an output layer, an activation function of which adopts ReLU; the input layer contains 12 nodes; the hidden layer is 2 layers, and each layer is provided with 32 nodes; and the output layer contains 6 nodes.

[0018] In the reference parameter library in step S1: the rotating speed response bandwidth of walking on flat ground at a uniform speed is 5Hz, and the torque upper limit threshold is 60% of the rated torque; the rotating speed response bandwidth of the running and jumping working condition is 20Hz, and the torque upper limit threshold is 120% of the rated torque; and the rotating speed response bandwidth of the climbing working condition is 10Hz, and the torque compensation coefficient is 1.3.

[0019] The dynamic correction in step S3 satisfies: when the load torque is greater than or equal to 80% of the rated torque, the PID proportion coefficient is increased by 15%; and when the motor temperature is greater than or equal to 60 DEG C, the torque upper limit threshold is reduced by 10%.

[0020] The load disturbance compensation in step S4 adopts a feedforward control algorithm, and the compensation response time is less than or equal to 5ms; the vibration suppression compensation adopts an adaptive notch filter algorithm, and the vibration attenuation rate is greater than or equal to 40%.

[0021] The protection logic in step S6 includes: when the motor current exceeds 1.5 times of the rated current for 30ms or the joint torque exceeds 1.8 times of the rated value, the motor output is automatically reduced, and a warning is sent.

[0022] Compared with the prior art, the present application has the following advantages and beneficial effects:

[0023] (1) The present application collects core parameters through multi-source sensor fusion, combines a "feature extraction + fuzzy neural network" combined algorithm, completes working condition recognition once every 5-10ms, has high recognition accuracy of 6 types of working conditions, and solves the problem of fuzzy working condition division caused by single parameter dependence in the prior art, and can accurately adapt to the differentiated control requirements of complex scenes such as flat ground, climbing and running and jumping.

[0024] (2) The present application is based on a preset reference parameter library and a real-time dynamic correction mechanism, optimizes the rotating speed response bandwidth, torque threshold and other core parameters for different working conditions, and simultaneously corrects through the load torque and the motor temperature linkage, so as to realize energy saving of walking on flat ground, high response of running and jumping, guarantee the load capacity of climbing, and balance energy saving, response speed and load performance.

[0025] (3) The present application adopts a triple dynamic compensation logic of feedforward control, adaptive notch filtering and mechanical characteristic compensation, and combines a closed-loop optimization mechanism, so that the joint motion angle error is less than or equal to 0.5 DEG, the joint resonance and load disturbance under high motion intensity are effectively inhibited, and the defects of large control error and obvious vibration in the prior art are solved.

[0026] (4) The present invention constructs a multi-dimensional protection logic for overload, overtemperature and sensor failure. When the motor current exceeds the rated value by 1.5 times (lasting for 30ms) or the joint torque exceeds the rated value by 1.8 times, the automatic load reduction warning is issued. When the sensor data is lost, the emergency mode is switched to avoid motor burnout or joint jamming, which greatly improves the operation safety and stability of the quadruped robot under complex working conditions. Attached Figure Description

[0027] Figure 1 This is a flowchart illustrating the control method of the present invention. Detailed Implementation

[0028] The present invention will be further described in detail below with reference to the embodiments, but the implementation of the present invention is not limited thereto.

[0029] Example

[0030] like Figure 1 As shown in the figure, the working condition adaptive joint module motor control method for quadruped robots described in this embodiment includes six steps, S1 to S6, as follows:

[0031] S1. After the system starts, it loads the preset reference parameter library for six working conditions: flat ground constant speed walking, flat ground acceleration / deceleration, climbing, descending, running and jumping, and obstacle crossing. It calibrates the zero point of the multi-source sensor and enters the "flat ground constant speed walking" working condition control mode by default.

[0032] This step establishes the system's initial configuration and working baseline, providing prerequisites for subsequent operating condition identification and parameter adaptation. Loading six operating condition baseline parameter libraries allows the system to quickly access corresponding control parameters; calibrating the zero point of multi-source sensors eliminates initial errors and ensures accurate data acquisition; and defaulting to the commonly used "flat ground constant speed walking operating condition control mode" ensures safe operation immediately after system startup without requiring additional manual settings.

[0033] The multi-source sensors include at least a joint angle sensor, a torque sensor, a motor current sensor, an IMU attitude sensor, and a six-dimensional force sensor. The joint angle sensor monitors the joint rotation angle and angular velocity, accurately feeding back the joint's motion posture and providing fundamental data for condition identification and control accuracy calibration. In this embodiment, the accuracy of the joint angle sensor is ≤0.1°.

[0034] Torque sensors are used to collect joint output torque in real time, directly reflecting the load size (such as increased torque when climbing a hill), and are the core basis for dynamic parameter correction and load capacity judgment.

[0035] Motor current sensor is used to monitor motor stator current, which indirectly reflects motor working load, and provides trigger signal for overload protection (current exceeds 1.5 times rated value) and working condition identification.

[0036] IMU attitude sensor is used to collect the pitch / roll / yaw attitude angle of the quadruped robot body, calculate the slope through the pitch angle, and identify the take-off state during running and jumping through the attitude change. In this embodiment, the sampling rate of the IMU attitude sensor is required to be ≥1000Hz.

[0037] Six-dimensional force sensor is used to detect ground contact pressure and multi-directional force, judge the contact state of the quadruped robot with the ground (such as pressure drop during running and jumping), assist in obstacle crossing, running and jumping and other working condition identification, and provide data for ground reaction force compensation. In this embodiment, the resolution of the six-dimensional force sensor is required to be ≤0.1N.

[0038] In the reference parameter library, for the flat ground walking working condition, the speed response bandwidth is required to be 5Hz, and the torque upper limit threshold is 60% of the rated torque; for the running and jumping working condition, the speed response bandwidth is required to be 20Hz, and the torque upper limit threshold is 120% of the rated torque; for the climbing working condition, the speed response bandwidth is required to be 10Hz, and the torque compensation coefficient is 1.3. Among them, the speed response bandwidth refers to the reaction speed of the motor to the control instruction, the higher the value, the more agile the motor adjusts the speed / torque. The torque upper limit threshold (percentage of rated torque) refers to the maximum torque upper limit allowed by the motor, which is used to prevent overload and adapt to the working condition requirements. The torque compensation coefficient refers to the additional force rule of the climbing working condition, that is, automatically increase 30% output (1.3 times) based on the reference torque, which is used to offset the gravity load during climbing to ensure sufficient climbing power.

[0039] S2, real-time collection of motor parameter data by multi-source working condition perception module, fusion analysis of the collected motor parameter data by combination algorithm, working condition identification every 5-10ms, and classification of the identified working condition into one of the above six working conditions.

[0040] The multi-source working condition perception module is a functional unit integrating hardware and algorithm, which takes multi-source sensors as the core hardware basis, and integrates feature extraction algorithm and fuzzy neural network algorithm, and is used for fusion analysis of the original data collected by the multi-source sensors, and finally completes the working condition identification.

[0041] The combination algorithm in the step is a set of feature extraction algorithm and fuzzy neural network algorithm, wherein the feature extraction algorithm is used to screen out core features capable of distinguishing different working conditions from 12 original data (joint angle, joint angular velocity, joint output torque, motor stator current, robot body pitch angle, robot body roll angle, robot body yaw angle, ground contact pressure X force, ground contact pressure Y force, ground contact pressure Z force, ground contact pressure X moment and ground contact pressure Y moment) collected by the multi-source sensor, reduce the high-dimensional original data to a low-dimensional effective feature vector, eliminate invalid data interference, highlight the working condition difference, and provide "precise materials" for subsequent classification.

[0042] The fuzzy neural network algorithm combines the advantages of fuzzy logic (good at dealing with fuzzy problems at the boundary of working conditions, such as slope critical value and uncertain state of running, jumping and taking off) and neural network (good at fitting complex mapping relationship and improving classification accuracy), receives the low-dimensional feature vector after feature extraction, and quickly determines that the current working condition belongs to a certain class of 6 core scenes through a preset network structure (12 inputs→2 hidden layers→6 outputs), and can meet the real-time requirement.

[0043] After the organic integration of the two, a synergistic link from "data preprocessing to precise classification" can be formed, efficient and accurate working condition recognition driven by multi-source data can be realized, and core basis can be provided for subsequent parameter self-adaptive adjustment and dynamic compensation, solving the limitation of single parameter working condition recognition in the prior art.

[0044] The fuzzy neural network algorithm has an input layer, a hidden layer and an output layer, and the activation function adopts ReLU; the input layer contains 12 nodes, and the 12 nodes correspond to the 12 original data, which are data inlets, and each node receives one data.

[0045] The hidden layer is 2 layers, and each layer is provided with 32 nodes. The hidden layer is used for data processing, and the 32 nodes correspond to 32 processing units, and two layers can deeply refine data features, such as distinguishing the "high torque + large pitch angle" feature of climbing.

[0046] The output layer contains 6 nodes, and the 6 nodes correspond to 6 working conditions, which are the outlet of the classification result, and each node outputs the matching probability of the corresponding working condition (such as determining the working condition when the probability of a certain node is the highest).

[0047] The ReLU activation function is used to enable the network to learn complex data relationships, while eliminating invalid signals to avoid processing delay.

[0048] The motor parameter data at least include joint angle / angle velocity, output torque, motor stator current, robot body attitude angle and ground contact pressure.

[0049] S3, if the system identifies the working condition switching, the reference parameters corresponding to the working condition are called, and the reference parameters are dynamically corrected based on the real-time collected motor parameters.

[0050] This step is used to realize accurate and dynamic parameter adaptation after working condition switching, link "working condition identification" and "dynamic compensation", guarantee the closed-loop effectiveness of control logic, avoid parameter mutation when working condition switching, realize smooth transition, and provide accurate parameter basis for subsequent dynamic compensation and closed-loop optimization.

[0051] Among them, the purpose of calling the reference parameters is to quickly match the core requirements of the new working condition (such as calling high response parameters for running and jumping, and calling high load parameters for climbing), and to avoid motion instability caused by control parameter mismatch after working condition switching. Dynamic correction of reference parameters is used to offset real-time load, temperature and other variable disturbances (such as sudden load increase and motor temperature rise), so that the control parameters are upgraded from "preset reference" to "real-time adaptation", and multiple target performance is considered.

[0052] The reference parameters at least include fuzzy adaptive PID parameter group, torque upper threshold, speed response bandwidth, and current protection threshold. The real-time collected motor parameters at least include load torque, motor temperature, and joint vibration acceleration data.

[0053] Among them, the fuzzy adaptive PID parameter group is an intelligent adjustment knob group adapted to the multi-working condition of quadruped robots, which is a combination of PID parameters and fuzzy logic adaptation. The PID parameters include proportional coefficient (Kp), integral coefficient (Ki), and differential coefficient (Kd), which cooperatively determine the motor control accuracy (such as angle error), response speed and stability. The fuzzy logic adaptation is not fixed like the fixed PID parameters, but can automatically fine-tune the proportional coefficient, integral coefficient and differential coefficient according to the real-time working condition (such as load change and vibration size) in combination with the fuzzy logic algorithm. The parameter group refers to a set of optimal reference PID parameters preset for 6 types of working conditions, which are stored in the reference parameter library and directly called and dynamically corrected when the working condition is switched.

[0054] The dynamic correction in this step meets the following conditions: when the load torque is greater than or equal to 80% of the rated torque, the PID proportional coefficient is increased by 15%; when the motor temperature is greater than or equal to 60℃, the torque upper threshold is reduced by 10%.

[0055] S4, output load disturbance compensation, vibration suppression compensation, and mechanical property compensation instructions through the dynamic compensation control module, and superimpose them into the motor control signal.

[0056] The purpose of this step is to offset the control deviation caused by load fluctuation, vibration interference and mechanical loss on the basis of parameter adaptation, make the motor control signal more accurate, ensure the joint motion accuracy and stability, and fill the gap between parameter adaptation and actual execution. Among them, the load disturbance compensation is used to compensate the sudden increase and decrease of load during the motion of the quadruped robot (such as sudden load increase when climbing and sudden load decrease after obstacle crossing). It captures the load change amount in real time through the torque sensor, generates compensation instructions using the feedforward control algorithm, and the compensation response time is ≤5ms to ensure that the motor output torque is adjusted in advance to avoid joint jamming or power shortage caused by sudden load change.

[0057] The vibration suppression compensation is used to compensate the joint resonance under high motion intensity such as running and jumping (such as joint vibration acceleration exceeding the standard when running and jumping). It generates reverse suppression instructions using adaptive notch filter algorithm based on IMU and joint vibration data, so as to ensure that the vibration attenuation rate is ≥40%, reduce the influence of resonance on control accuracy, and avoid wear of motor and mechanical structure.

[0058] The mechanical property compensation is used to compensate the mechanical loss of joint module (such as reduction ratio drift, transmission efficiency decline and inertia change). It pre-stores the mechanical property curve and corrects the control instruction according to the real-time speed and load data to eliminate the system error caused by mechanical aging or working condition change, and ensure that the motor output is accurately matched with the actual demand of joint.

[0059] S5, based on the joint angle and motor current feedback data, continuously adjust the control parameters to ensure that the joint motion angle error is ≤0.5°.

[0060] This step is the closed-loop optimization link of the embodiment, which aims to correct the deviation in real time and lock the control accuracy. Among them, the joint angle feedback data comes from the joint angle sensor (accuracy ≤0.1°), and the motor current feedback data comes from the motor current sensor.

[0061] The object of adjusting the control parameters described herein includes: fuzzy adaptive PID parameter group and auxiliary adjustment of speed response bandwidth and torque output accuracy. Among them, for the fuzzy adaptive PID parameter group (proportion coefficient Kp, integral coefficient Ki and differential coefficient Kd), when the angle deviation is large, the proportion coefficient Kp is increased to speed up the correction; when the deviation persists, the integral coefficient Ki is increased to eliminate the static error. For auxiliary adjustment of speed response bandwidth and torque output accuracy, if the current feedback shows that the load is too high, the bandwidth is fine-tuned to ensure the balance between power and accuracy.

[0062] The system synchronously collects feedback data every 5ms, calculates the angle deviation in real time; according to the deviation size, deviation trend and motor current load, dynamically fine-tune the control parameters to form a cycle optimization of "collection-comparison-calculation-adjustment", until the angle deviation is stable within ≤0.5°.

[0063] S6, real-time monitoring of motor current, joint torque and motor temperature, when overload, over-temperature or sensor failure is triggered, protection logic is executed and a warning signal is sent.

[0064] This step is a safety protection and fault warning link, the core of which is to avoid motor burnout due to overload and over-temperature, joint stuck due to torque overrun, and reduce the damage risk of mechanical structure and electronic components. Among them, for the monitoring of motor current, it is real-time tracking of motor stator current, and the trigger condition is that the current exceeds 1.5 times of the rated current for 30 ms; for the monitoring of joint torque, it is tracking the output torque of the joint, and the trigger condition is that the torque exceeds 1.8 times of the rated value, so as to avoid the fracture of joint gear set and transmission shaft due to overload; for the monitoring of motor temperature, it is monitoring the temperature of motor shell, and the trigger condition is that the temperature is greater than or equal to 60 DEG C; for the monitoring of sensor failure, it is to check the validity of multi-source sensor data synchronously, and the trigger condition is that there is no valid data for a certain type of sensor for 10 ms, such as angle sensor zero drift anomaly, data loss, etc.

[0065] When the trigger condition is reached, protection logic is executed and a warning signal is sent. The protection logic is that when overload / over-temperature occurs, the motor output is automatically reduced, i.e. the torque upper limit is reduced to 50% of the rated value and the speed response bandwidth is reduced to 5 Hz, until the parameters return to the safe range, and if the overrun continues, the machine is stopped; when the sensor fails, the emergency working condition is switched immediately, and the preset conservative parameters (low response, low load) are loaded to ensure that the quadruped robot slows down or stops smoothly, and to avoid control misalignment due to data loss.

[0066] The sending object of the warning signal is the main control system of the quadruped robot and the remote operator terminal, and the signal content at least includes the fault type (such as "motor overload", "temperature overrun", "torque sensor failure") and real-time monitoring data (such as current / temperature / torque value), so as to facilitate quick positioning of the problem.

[0067] As described above, the application can be well implemented.

Claims

1. A working condition adaptive joint module motor control method for quadruped robots, characterized in that, Includes the following steps: S1. After the system starts, it loads the preset reference parameter library for six working conditions: flat ground constant speed walking, flat ground acceleration / deceleration, climbing, descending, running and jumping and obstacle crossing, calibrates the zero point of the multi-source sensor, and enters the "flat ground constant speed walking" working condition control mode by default. S2. The motor parameter data is collected in real time through the multi-source working condition sensing module. The collected motor parameter data is fused and analyzed by the combined algorithm. The working condition is identified every 5 to 10 ms, and the identified working condition is divided into one of the above six working conditions. S3. If the system detects a change in operating condition, it calls the reference parameters for the corresponding operating condition and dynamically corrects the reference parameters based on the real-time collected motor parameters. S4. Output load disturbance compensation, vibration suppression compensation, and mechanical characteristic compensation commands through the dynamic compensation control module and superimpose them into the motor control signal; S5. Based on joint angle and motor current feedback data, continuously adjust control parameters to ensure that the joint motion angle error is ≤0.5°; S6. Monitor motor current, joint torque and motor temperature in real time. When overload, overtemperature or sensor failure is triggered, execute protection logic and send a warning signal.

2. The working condition adaptive joint module motor control method for quadruped robots according to claim 1, characterized in that, The "multi-source sensors" mentioned in step S1 include at least a joint angle sensor, a torque sensor, a motor current sensor, an IMU attitude sensor, and a six-dimensional force sensor; the "motor parameter data" mentioned in step S2 includes at least joint angle / angular velocity, output torque, motor stator current, robot body attitude angle, and ground contact pressure.

3. The working condition adaptive joint module motor control method for quadruped robots according to claim 2, characterized in that, The "baseline parameters" mentioned in step S3 include at least the fuzzy adaptive PID parameter group, torque upper limit threshold, speed response bandwidth, and current protection threshold; the "real-time acquired motor parameters" include at least the load torque, motor temperature, and joint vibration acceleration data.

4. The working condition adaptive joint module motor control method for quadruped robots according to claim 2, characterized in that, The "combined algorithm" mentioned in step S2 is a feature extraction algorithm and a fuzzy neural network algorithm.

5. A working condition adaptive joint module motor control method for quadruped robots according to claim 2, characterized in that, In the multi-source working condition sensing module, the accuracy of the joint angle sensor is ≤0.1°, the sampling rate of the IMU attitude sensor is ≥1000Hz, and the resolution of the six-dimensional force sensor is ≤0.1N.

6. The method for adaptive joint module motor control for quadruped robots according to claim 4, characterized in that, The fuzzy neural network algorithm has an input layer, a hidden layer, and an output layer, and its activation function is ReLU; the input layer has 12 nodes; the hidden layer has 2 layers, and each layer has 32 nodes; the output layer has 6 nodes.

7. A working condition adaptive joint module motor control method for quadruped robots according to any one of claims 1 to 6, characterized in that, In the reference parameter library mentioned in step S1: the speed response bandwidth for flat, uniform walking is 5Hz and the upper limit threshold of torque is 60% of the rated torque; the speed response bandwidth for running and jumping conditions is 20Hz and the upper limit threshold of torque is 120% of the rated torque; the speed response bandwidth for climbing conditions is 10Hz and the torque compensation coefficient is 1.

3.

8. A working condition adaptive joint module motor control method for quadruped robots according to claim 7, characterized in that, The dynamic correction described in step S3 satisfies the following: when the load torque is ≥ 80% of the rated torque, the PID proportional coefficient increases by 15%; when the motor temperature is ≥ 60℃, the upper limit threshold of torque decreases by 10%.

9. A working condition adaptive joint module motor control method for quadruped robots according to claim 8, characterized in that, The load disturbance compensation in step S4 adopts a feedforward control algorithm with a compensation response time ≤ 5ms; the vibration suppression compensation adopts an adaptive notch filter algorithm with a vibration attenuation rate ≥ 40%.

10. A working condition adaptive joint module motor control method for quadruped robots according to claim 9, characterized in that, The protection logic described in step S6 includes: when the motor current exceeds 1.5 times the rated current or the joint torque exceeds 1.8 times the rated value for 30ms, the motor output is automatically reduced and an early warning is sent.

Images