Slip suppression control method and system for positioning and pose adjustment of four-wheel independent steering AGV

Abstract

The application discloses a four-wheel independent rudder wheel AGV positioning and pose adjusting slip suppression control method and system, the method comprises the following steps: monitoring that the AGV enters a preset accurate positioning area, and switching to a slip suppression control mode; a Kalman filter is constructed, real vehicle body motion state quantities without slip interference are output, and the slip rates of each rudder wheel are calculated; taking the current AGV total pose error and the real-time slip rate as input, reasoning is carried out according to a preset fuzzy rule, and the current control step output is generated; combined with the current control step output, taking the slip rate as a constraint input, the driving torque of each rudder wheel is calculated by using a compound control law; based on the instantaneous rotation center principle, the inverse kinematics is used to solve the four-wheel steering angle and the wheel speed; based on the current control step output, a subsequent stop calibration stage is entered, and three-degree-of-freedom pose deviation recalculation is carried out during the stop period; the step-by-step reversing and calibration are circularly carried out until the positioning accuracy requirement is reached. The application can improve the AGV parking accurate positioning accuracy.

Description

Technical Field

[0001] This application relates to the field of positioning and attitude adjustment of automated guided vehicles and slip suppression control, specifically to a slip suppression control method and system for positioning and attitude adjustment of a four-wheeled independent steering wheel AGV. Background Technology

[0002] In the field of industrial automation, Automated Guided Vehicle (AGV) technology has developed rapidly, playing a vital role in material handling, warehousing and logistics, and other scenarios, greatly improving production efficiency and reducing labor costs. In particular, four-wheeled independent steering wheel AGVs, due to their flexible movement and steering capabilities, are widely used in complex industrial environments and can adapt to different work scenarios and task requirements.

[0003] When performing precise positioning and attitude correction operations on four-wheeled independent steering wheel AGVs, several common technical methods have been employed. Some AGVs use open-loop control for reversing operations, which executes actions based on a pre-set program without considering changes in the actual operating state. Other AGVs utilize a single PID closed-loop control, adjusting for deviations through a feedback system, but the control logic is relatively simple. Additionally, some anti-skid solutions directly borrow from automotive snow driving modes or ESP control logic, attempting to enhance the AGV's anti-skid capabilities.

[0004] Clearly, existing technologies have significant shortcomings. In multi-degree-of-freedom fine-tuning scenarios, the friction state between the polyurethane wheel and the smooth surface is easily altered when the steering wheel reverses direction in place. Traditional solutions, relying solely on encoder feedback, struggle to accurately identify wheel spin and vehicle slippage, leading to erroneous posture signals. Furthermore, traditional torque control lacks inertia and friction feedforward, resulting in large torque impacts during reversals, exacerbating slippage. Attitude adjustments often employ one-time, large-angle reversals without dynamic adjustments based on slippage conditions, leading to poor coordination between four-wheel steering and speed, and susceptibility to lateral slippage due to motion interference. Moreover, anti-slip solutions that partially borrow from automotive control logic fail to fully consider the characteristics of independent four-wheel drive and limited-range in-place posture adjustment. They cannot achieve robust real-time slippage detection, smooth torque output, and coordinated control through dynamic reversal planning, making it difficult to balance precise positioning accuracy and posture stability, and ultimately failing to meet the industrial demands of high-precision parking of multiple AGVs and precise production line docking. Summary of the Invention

[0005] To achieve coordinated control of AGV slippage real-time robust detection, smooth torque output, and dynamic reversing planning, and to significantly improve the precise positioning accuracy of AGV parking, this application provides a slippage suppression control method and system for positioning and attitude adjustment of a four-wheel independent steering wheel AGV.

[0006] In a first aspect, this application provides a slip suppression control method for positioning and attitude adjustment of a four-wheeled independent steering wheel AGV, including: S1. Monitor the AGV to enter the preset fine positioning area and start the fine positioning and posture adjustment process, then switch to the slip suppression control mode. S2. Collect angular velocity data from the encoders of each steering wheel, IMU data, and laser positioning data; construct a Kalman filter with vehicle coordinates, heading angle, translational velocity, and angular velocity as state variables; iteratively calculate the state update equation and the observation equations of the fused encoder, IMU, and laser positioning to output the true vehicle motion state variables without slippage interference; calculate the slip ratio of each steering wheel based on the true vehicle motion state variables to determine the total pose error of the AGV; S3. Using the current total AGV pose error and real-time slip rate as input, reason according to the preset fuzzy rules to generate the single-step expected reversing angle of the current control step, the maximum safety limit of the driving torque of each wheel, and the pause observation time after the reversing action is executed. S4. Combining the current control step output with the slip ratio as the constraint input, calculate the driving torque of each steering wheel using a composite control law; the composite control law includes: a PID closed-loop term based on real speed feedback, a moment of inertia feedforward term, a viscous and Coulomb friction compensation term, a ground load friction feedforward term, and a slip limit term that dynamically calculates the upper limit of torque based on the current slip ratio and a preset suppression coefficient. S5. Based on the current control step output, and using the principle of instantaneous rotation center, solve the steering angle and wheel speed of the four wheels using inverse kinematics to ensure that the speed vectors of the four wheels converge at the same instantaneous rotation center. S6. Based on the obtained driving torque, steering angle and wheel speed of each steering wheel after amplitude limiting, perform single-step reversal, and then enter the pause calibration stage. During the pause, perform pose three-degree-of-freedom deviation recalibration calculation. S7. Determine whether the pose deviation meets the preset precision positioning accuracy requirements; if not, return to step S3 to continue iteration; if it meets the requirements, exit the pose adjustment mode.

[0007] By adopting the above scheme, after the AGV enters the precision positioning area, the actual slippage of the vehicle body can be accurately identified by Kalman filtering. Based on fuzzy rules, the reversing parameters are dynamically planned, and the torque is smoothly output through a composite control law. Combined with four-wheel cooperative constraints, motion interference is eliminated. Through the reversing, pausing, and calibration cycle operation, slippage suppression is achieved during the AGV positioning and attitude adjustment process, which greatly improves the precision positioning accuracy of parking.

[0008] Preferably, the calculation of the slip ratio of each steering wheel in step S2 includes: The actual vehicle speed obtained from the filtering solution is converted into the longitudinal and lateral speeds in the coordinate system of each steering wheel center. Combined with the measured wheel speed and steering angle of each steering wheel, the initial value of the kinematic slip ratio of each steering wheel is calculated. The real-time output torque of each steering wheel drive motor and steering motor is collected, and the longitudinal and lateral forces at the wheel ends are calculated. Combined with the vertical load of a single wheel, the real-time adhesion utilization rate of the current steering wheel is calculated. A recursive least squares algorithm with a forgetting factor is used, with the initial value of the kinematic slip ratio and the real-time adhesion utilization rate as input, to identify the peak adhesion coefficient and the optimal slip ratio parameter corresponding to the current road surface online, and update the constructed steering wheel slip-adhesion model; the steering wheel slip-adhesion model is a model describing the relationship between the current slip ratio and the peak adhesion coefficient of the road surface. Based on the updated steering wheel slip-adhesion model, the dynamic correction slip ratio is obtained by inverting the real-time adhesion utilization rate. Then, the initial value of the kinematic slip ratio and the dynamic correction slip ratio are adaptively weighted and fused to output the final real-time slip ratio of each steering wheel.

[0009] By adopting the above scheme and taking into account kinematic and dynamic factors, the real-time slip ratio of each steering wheel can be calculated more accurately, providing more reliable data support for subsequent slip suppression control, thereby better realizing slip suppression during the AGV positioning and attitude adjustment process.

[0010] Preferably, step S3 further includes: Calculate the asymmetry of the AGV's left and right side slippage in real time and determine the global maximum slippage rate; Using the current AGV's total pose error, global maximum slip ratio, and left-right slip asymmetry as fuzzy inputs, fuzzy inference yields the desired yaw rate ratio, desired translational speed ratio, and instantaneous steering center ICR offset factor. Based on the fuzzy inference output, the expected single-step reversal angle, the maximum safety limit of each wheel drive torque, and the pause observation duration after the reversal action are calculated for the current control step. The expected single-step reversal angle is obtained by multiplying the expected yaw rate ratio by the preset maximum single-step rotation angle; the maximum safety limit of each wheel drive torque is based on... The determination is based on the smaller of the desired yaw rate ratio and the desired translational speed ratio, the preset rated torque, and a reduction coefficient determined by the real-time slip risk. The pause observation duration is set inversely based on the combined motion intensity of the desired yaw rate ratio and the desired translational speed ratio; the lower the motion intensity and the higher the slip risk, the longer the pause duration. The real-time slip risk is determined by quantifying the slip risk index and calculating the slip risk index value of each wheel based on the real-time slip rate. The ICR offset factor adjusts and corrects the instantaneous steering center position under multi-wheel slip asymmetry conditions. The fuzzy inference rule design includes: the inputs are the total pose error of the AGV, the global maximum slip ratio, and the left and right slip asymmetry, and each input parameter is set with several levels; the outputs are the expected yaw rate ratio, the expected translational rate ratio, and the instantaneous steering center ICR offset factor, and each output parameter is combined with the preset input parameters to form a mapping.

[0011] By adopting the above scheme, taking into account factors such as the total pose error of the AGV, the real-time slip rate, and the left and right side slip asymmetry, the scheme uses fuzzy inference to dynamically generate the single-step expected reversal angle, the maximum safety limit of the driving torque of each wheel, and the pause observation time. This better adapts to the multi-wheel slip asymmetry working condition, ensures that the target rotation angle of each wheel meets the Ackerman cooperative kinematics constraint, and further improves the slip suppression effect and fine positioning accuracy in the AGV positioning and posture adjustment process.

[0012] Preferably, step S3 further includes: Construct a deep reinforcement learning network; the construction steps include: defining the state as the total pose error of the front AGV and the real-time slip rate; defining the action as a continuous action, controlling the single-step expected reversal angle, the maximum safety limit of the driving torque of each wheel, and the pause observation time after the reversal action; defining the reward function based on the weighted calculation of the arrival positioning distance error reward and the slip penalty; Quantify the slip risk indicators, calculate the slip risk indicator values ​​for each round based on the real-time slip rate, and combine the slip risk indicator values ​​for each round to generate a weighted global slip risk index; When the global slip risk index is greater than the first threshold, inference is performed based on the preset fuzzy rules, and the expected single-step reversal angle, maximum safety limit, and pause observation duration are output. When the global slip risk index is less than the second threshold, the expected single-step reversal angle, maximum safety limit, and pause observation duration are calculated using a deep reinforcement learning network. When the global slip risk index is between the first and second thresholds, the output of the preset fuzzy rules and the output of the deep reinforcement learning network are weighted and fused to obtain the final output.

[0013] By adopting the above scheme, combining deep reinforcement learning networks and preset fuzzy rules, and adaptively selecting control strategies based on the global slip risk index, the single-step expected reversal angle, maximum safety limit, and pause observation duration are generated more accurately, effectively balancing attitude adjustment efficiency and accuracy, and further suppressing slippage during AGV positioning and attitude adjustment.

[0014] Preferably, the composite control law in step S4 further includes: The inverse dynamic feedforward compensation term is based on the single-step expected commutation angular acceleration and the vehicle's rotational inertia. This compensation term is calculated based on the single-step expected commutation angular acceleration corresponding to the differential of the generated single-step expected commutation angular velocity, combined with the vehicle's rotational inertia about the vertical axis and the geometric layout parameters of the vehicle body and steering wheel. The additional driving torque required to overcome the vehicle's inertial torque is then added to the total torque of each wheel as a feedforward quantity to pre-counteract the load changes caused by dynamic motion.

[0015] By adopting the above scheme, the additional driving torque required to overcome the inertial torque of the whole vehicle is calculated and superimposed on the total torque of each wheel, thus pre-counteracting the load changes caused by dynamic motion, making the composite control law more perfect, and improving the accuracy and stability of positioning and attitude adjustment.

[0016] Preferably, the pose three-degree-of-freedom deviation recalibration calculation performed during the pause in step S6 further includes: When generating the output of the current control step, the theoretical motion trajectory under ideal no-slip conditions is simultaneously calculated based on the kinematic model of the AGV and the current starting pose, and used as a reference trajectory. During the pause calibration phase, the actual motion trajectory is obtained by fusing multi-source sensor data through a Kalman filter, and the actual motion trajectory is dynamically compared with the reference trajectory to obtain the differences in trajectory shape, convergence speed and endpoint pose. Based on the differences, the sources of deviation are traced and distinguished, and the sources of deviation are distinguished as inaccurate kinematic parameters, dynamic response lag, or local slip events, and corresponding compensation parameters are generated. The corresponding compensation parameters are then used to correct the AGV kinematic model parameters, composite control law parameters, and Kalman filter parameters online.

[0017] By adopting the above scheme, the source of the pose deviation is traced, compensation parameters are generated, and relevant parameters are corrected online, thereby reducing positioning errors caused by inaccurate kinematic parameters, lag in dynamic response, or local slippage events, and improving the accuracy and stability of AGV positioning and posture adjustment.

[0018] Secondly, this application provides a slip suppression control system for positioning and attitude adjustment of a four-wheeled independent steering wheel AGV, comprising: The mode switching module is used to monitor the AGV entering the preset fine positioning area and start the fine positioning and posture adjustment process, switching to the slip suppression control mode; The slip detection module is used to collect angular velocity data from the encoders, IMU data, and laser positioning data of each steering wheel; it constructs a Kalman filter with vehicle coordinates, heading angle, translational velocity, and angular velocity as state variables, and iteratively calculates the state update equation and the observation equations of the fused encoder, IMU, and laser positioning to output the true vehicle motion state variables without slip interference; based on the true vehicle motion state variables, it calculates the slip ratio of each steering wheel to determine the total pose error of the AGV; The reversing control module is used to take the current AGV total pose error and real-time slip rate as input, and perform inference according to preset fuzzy rules to generate the single-step expected reversing angle of the current control step, the maximum safety limit of the driving torque of each wheel, and the pause observation time after the reversing action is executed. The torque control module is used to combine the output of the current control step with the slip ratio as the constraint input and to calculate the driving torque of each steering wheel using a composite control law. The composite control law includes: a PID closed-loop term based on real speed feedback, a moment of inertia feedforward term, a viscous and Coulomb friction compensation term, a ground load friction feedforward term, and a slip limit term that dynamically calculates the upper limit of torque based on the current slip ratio and a preset suppression coefficient. The motion coordination module is used to solve the steering angle and wheel speed of the four wheels based on the current control step output and the principle of instantaneous rotation center, using inverse kinematics to ensure that the speed vectors of the four wheels converge at the same instantaneous rotation center. The suppression execution module is used to perform single-step reversal based on the acquired drive torque, steering angle and wheel speed after each steering wheel is limited, and then enter the pause calibration stage. During the pause, the pose three-degree-of-freedom deviation recalibration calculation is performed; it is determined whether the pose deviation meets the preset fine positioning accuracy requirements; if not, it returns to the slip detection module to repeat the reversal control, torque control and cooperative motion until the requirements are met, and then exits the attitude adjustment mode.

[0019] By adopting the above scheme, slip suppression control is activated in a timely manner when the AGV enters the precision positioning area. The slip rate of each steering wheel is accurately detected, the reversing parameters are adaptively planned, and the smooth drive torque is output to ensure the coordinated movement of the four wheels. The position and posture deviation is corrected when performing single-step reversing and pause calibration, and finally slip suppression is achieved in the AGV positioning and posture adjustment process, which effectively improves the precision positioning accuracy of parking.

[0020] Thirdly, this application provides a computer-readable storage medium including a stored computer program, wherein the computer program, when running, controls the device where the computer-readable storage medium is located to perform the method described above.

[0021] Fourthly, this application provides a computer device, the computer device including a memory, a processor and a program stored in the memory and executable thereon, the program being executed by the processor to implement the steps of the method described above.

[0022] In summary, this application has the following beneficial effects: 1. By monitoring the AGV entering the preset precision positioning area and initiating the precision positioning and attitude adjustment process, the system switches to slip suppression control mode. Multi-source data is collected to construct a Kalman filter, outputting the true vehicle motion state without slip interference and calculating the slip ratio of each steering wheel. Control step parameters are generated based on preset fuzzy rules, and the driving torque is calculated using a composite control law. The steering angle and wheel speed of the four wheels are solved based on the instantaneous rotation center principle. Single-step reversal is performed with pauses and calibrations, iterating continuously until the precision positioning accuracy requirements are met. This achieves slip suppression during the AGV positioning and attitude adjustment process, improving positioning accuracy and attitude adjustment stability. 2. In reversing control, the asymmetry of left and right side slippage of the AGV and the global maximum slippage rate are calculated. Multiple parameters are used as fuzzy inputs for inference to generate accurate control step parameters. A deep reinforcement learning network is constructed to select different output methods according to the global slippage risk index to achieve more accurate fuzzy distribution output. Based on these methods, different slippage conditions are better adapted to, the reversing and motion coordination strategies are optimized, slippage is effectively suppressed, and the performance of AGV positioning and attitude adjustment is improved. 3. When calculating the slip ratio of each steering wheel, kinematic and dynamic factors are comprehensively considered, the steering wheel slip-adhesion model is updated, and adaptive weighted fusion is used to obtain the final slip ratio, making the slip ratio calculation more accurate; an inverse dynamic feedforward compensation term based on the single-step expected directional acceleration and the vehicle's rotational inertia is added to the composite control law to pre-counterload load changes caused by dynamic motion; during the pause calibration stage, the actual motion trajectory is compared with the reference trajectory, the source of deviation is traced and identified, and compensation parameters are generated to correct relevant parameters, further improving the accuracy and reliability of positioning and attitude adjustment. Attached Figure Description

[0023] Figure 1 This is a flowchart of the slip suppression control method for positioning and attitude adjustment of a four-wheeled independent steering wheel AGV as described in a specific embodiment; Figure 2 This is a schematic diagram of the slip suppression control system for the positioning and attitude adjustment of the four-wheeled independent steering wheel AGV described in a specific embodiment. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0025] like Figure 1 As shown in the figure, this application discloses a slippage suppression control method for positioning and attitude adjustment of a four-wheeled independent steering wheel AGV, including: monitoring and initiation, data acquisition and slippage rate calculation, reversal parameter generation, torque calculation, motion cooperative solution, execution and calibration, and exit judgment. Each step is executed sequentially to form a closed-loop control process. Through real-time monitoring and dynamic adjustment, slippage of the AGV during positioning and attitude adjustment is effectively suppressed, improving positioning accuracy. The following is a further detailed description of this application.

[0026] Specifically, in the monitoring startup step, sensors are used to monitor whether the AGV has entered the preset precision positioning area. When the sensor detects that the AGV has entered the preset precision positioning area, the precision positioning and attitude adjustment process is automatically initiated, switching to the slip suppression precision positioning mode. The sensors can be laser sensors, infrared sensors, etc., to sense the AGV's position information in real time to determine whether it has entered the preset precision positioning area.

[0027] S2. Collect multi-source sensing data and use Kalman filtering to calculate the vehicle's true pose, speed and slip ratio in real time.

[0028] Specifically, when switching to the slip suppression control mode, step S2 is executed. The multi-source sensing acquisition step includes acquiring the angular velocity of each steering wheel encoder, IMU data, and laser positioning data. The encoders accurately measure the rotational speed of the steering wheels, the IMU provides the AGV's attitude information, and the laser positioning data determines the AGV's position.

[0029] In the real-time solution process using Kalman filtering, a Kalman filter is pre-constructed. The construction process includes: using vehicle coordinates, heading angle, translational velocity, and angular velocity as state variables; and selecting the system state variable formula as follows: In the formula, Here are the vehicle's position coordinates. For heading angle, The average speed of the vehicle body. The angular velocity is used; the state update equation updates the state variables based on the dynamic characteristics; the formula is: The observation equation integrates observations from the encoder, IMU, and laser positioning; the observation equation integrates multi-source signals, and the formula is: In the formula, This is a fusion of observations from laser positioning, IMU, and encoder. , This refers to process and observation noise.

[0030] The system outputs the true vehicle motion state parameters without slippage interference through filtering and iteration; based on the true vehicle motion state parameters, the slip ratio of each steering wheel is calculated to determine the total AGV pose error. The formula for calculating the longitudinal slip ratio of a single steering wheel is as follows: In the formula, Let i be the slip ratio of the i-th steering wheel. The encoder angular velocity is r, and the effective radius of the polyurethane wheel is r. This represents the actual speed of the vehicle along the wheel direction. It can be set when | When |>0.01, it is considered a valid slip.

[0031] Thus, the true position and orientation of the vehicle body are obtained. ), vehicle speed ( and the slip ratio of each wheel. (This embodiment focuses on longitudinal slip ratio).

[0032] S3. Based on the parameters calculated in real time, the fuzzy controller outputs step-by-step commutation parameters and torque limits to achieve gradual attitude adjustment.

[0033] Specifically, in the step-by-step reversing parameter generation step, the current AGV total pose error and real-time slip rate are used as inputs, and reasoning is performed according to the preset fuzzy rules corresponding to the fuzzy controller to obtain the reversing parameters and torque limit of the current control step.

[0034] The input variable of the fuzzy controller is the positioning error between the true pose of the vehicle and the target pose of the vehicle. To positioning error , Heading angle error and based on each The weighted average is used to calculate the real-time slip ratio S.

[0035] Among them, the output variable of the fuzzy controller is the single-step expected reversal angle Δ. Maximum safety limits for torque of each drive wheel and the duration of pause observation (calibration) after performing a reversing action. .

[0036] The fuzzy control rules of the fuzzy controller can be set as follows: the real-time slip rate of the input variables can be set to multiple different levels such as high and low, and the pose error can be set to different levels such as large and small; corresponding output adjustment rules are set for different combinations of levels of each variable, and the output variables are obtained based on the preset basic parameter values ​​and the adjustment rules. For example: when the real-time slip rate is high and the pose error is small, the adjustment rule is: reduce the reversal step size, reduce the torque, and extend the pause time; when the real-time slip rate is low and the pose error is large, the adjustment rule is: use a medium step size and normal torque to improve the pose adjustment efficiency; when the real-time slip rate is low and the pose error is small, close to the target accuracy, the adjustment rule is: enable micro-step reversal to ensure positioning accuracy.

[0037] S4. Combining the current control step output with slip ratio as the constraint input, calculate the driving torque of each steering wheel using a composite control law.

[0038] Specifically, for the current control step output planned by the fuzzy controller, a combination of PID and feedforward control is used with slip ratio as a constraint to achieve smooth torque output and avoid commutation shock and slippage.

[0039] The composite control law includes: a PID closed-loop term based on real speed feedback, a moment of inertia feedforward term, a viscous and Coulomb friction compensation term, a ground load friction feedforward term, and a slip limiting term that dynamically calculates the upper limit of torque based on the current slip ratio and a preset suppression coefficient; the formula is: ;in, The formula for the PID closed-loop term based on real speed feedback is: ; In the formula, The true velocity output by the Kalman filter is obtained, eliminating the velocity error caused by slippage. All are corresponding coefficients. Among them, The feedforward compensation term is given by the following formula: In the formula, For rotational inertia feedforward, f(ω) is the viscous and Coulomb friction compensation. This is a ground load friction feedforward mechanism to proactively counteract disturbances. Among other things, The slip limit term is calculated using the following formula: In the formula, The slip ratio is the slip suppression coefficient. The larger the slip ratio, the more stringent the torque limitation, thus suppressing slippage at its source. This is the maximum safety limit.

[0040] S5. Based on the current control step output, and using the principle of instantaneous rotation center, solve the steering angle and wheel speed of the four wheels using inverse kinematics to ensure that the speed vectors of the four wheels converge at the same instantaneous rotation center.

[0041] Specifically, in order to prevent slippage caused by interference of four-wheel motion, a collaborative model is established based on the instantaneous rotation center (ICR) to ensure that the velocity vectors of the four wheels converge at the same rotation center.

[0042] Among them, based on the single-step expected reversal angle Δ Maximum safety limits for torque of each drive wheel and the duration of pause observation (calibration) after performing a reversing action. Calculation obtained The total cycle time for a single step is divided into two parts: one part is the reversal execution time. (Time for wheel rotation and vehicle body to complete pose increment) and pause calibration time (Fuzzy output quantity); Preset engineering parameters (typically 0.2-1s, smaller step size results in shorter execution time) can be used, or fuzzy rules can be used for synchronous output; the higher the slip ratio, The longer the length, the lower the acceleration and the less impact slip. In the formula, , Let be the instantaneous velocity component of the center of the i-th steering wheel; The fixed coordinates of the center of the i-th steering wheel can be determined based on sensor data and mechanical structure; The coordinates of the instantaneous rotation center (ICR); , This refers to the translational velocity component of the ICR point itself; where, in a pure heading correction in-situ attitude adjustment scenario, the ICR coincides with the vehicle's center of mass (or can be specified as any point, such as the geometric center, as required). And under pure rotation .

[0043] Furthermore, if a composite attitude adjustment with translation is used, the ICR coordinates can be derived from the relationship between translational velocity and angular velocity. In step S4, the single-step output not only outputs the desired single-step reversal angle Δ It can also provide supplementary output: the single-step output also includes translational corrections Δ in the X / Y directions. Δ The values ​​can be proportionally allocated from the total pose error, or output using fuzzy rules, with logic completely consistent with the rotation angle. Assuming the vehicle body performs uniform translation and uniform rotation within a single step, the instantaneous velocity of the center of mass and the yaw rate are: ; Any motion of the corresponding planar rigid body can be equivalent to a pure rotation about a fixed point (ICR); Together, they determine the coordinates of the ICR in the vehicle body coordinate system: , Final calculation: , In the formula, Let i be the steering angle of the i-th steering wheel. Let be the speed of the i-th steering wheel. Real-time inverse kinematics is used to ensure that the steering and speed of the four wheels are synchronized, thus eliminating internal stress slippage.

[0044] S6. Based on the acquired driving torque, steering angle and wheel speed after each steering wheel is limited, perform a single-step reversal, and then enter the pause calibration stage. During the pause, perform pose three-degree-of-freedom deviation recalibration calculation.

[0045] Specifically, after a single-step reversal is completed, the system pauses for a period of time set by the fuzzy controller, primarily to allow mechanical vibration to decay. During this pause, high-precision positioning sensors (such as laser positioners and vision systems) are used to measure multi-source data and obtain the true absolute pose of the vehicle body. The current pose after this step is compared and calibrated to obtain the steering operation execution error for this step, thereby calculating the new residual error and completing the execution of the X / Y pose. Three-degree-of-freedom deviation recalibration calculation is performed to facilitate subsequent error correction.

[0046] S7. Determine whether the pose deviation meets the preset precision positioning accuracy requirements.

[0047] Specifically, determine whether the remaining total error after calibration meets the preset positioning accuracy requirements, such as... , If the conditions are met, the pose adjustment is complete, and the process proceeds to step S8; if not, the process jumps to step S3, where the fuzzy controller plans the parameters for the next step based on the updated current pose (after calibration), the new remaining total error, and the real-time slip rate. Then, steps S4, S5, and S6 are executed until the preset positioning accuracy requirement is met.

[0048] S8, Slip Inhibition Mode.

[0049] By adopting the above scheme, the true motion state and slip ratio of the AGV are obtained based on multi-sensor data fusion and Kalman filtering. A combination of fuzzy control and deep reinforcement learning is used to dynamically adjust the reversing parameters according to the pose error and slip state, achieving gradual attitude adjustment in small steps. Feedforward compensation and slip limiting in the composite control law are used to suppress torque impact and slippage. Motion coordination constraints based on the instantaneous rotation center eliminate slippage caused by interference from the four wheels. Trajectory comparison and parameter correction during the pause calibration phase promptly correct deviations and ensure positioning accuracy. The entire control process forms a closed loop, effectively suppressing slippage of the AGV during positioning and attitude adjustment, improving positioning accuracy and attitude adjustment stability, and meeting the industrial needs of high-precision AGV parking and precise production line docking.

[0050] In one specific embodiment, considering that the traditional slip ratio is calculated solely by the kinematic ratio of wheel speed and vehicle speed, which is susceptible to speed estimation errors, insufficient resolution of low-speed encoders, and significant sensor noise, a method is adopted to more accurately calculate the slip ratio. This method employs an approach combining initial kinematic values, dynamic constraints, and joint identification and inversion. The slip ratio is corrected using the physical constraints of the road adhesion coefficient, achieving synchronous high-precision estimation. The method further includes: step S2, which involves calculating the slip ratio of each steering wheel, including: First, the actual vehicle speed obtained from the filtering solution is converted into the longitudinal and lateral speeds in the coordinate system of each steering wheel center. Combined with the measured wheel speed and steering angle of each steering wheel, the initial value of the kinematic slip ratio of each steering wheel is calculated.

[0051] Calculate the longitudinal slip ratio of a single steering wheel: ; Calculate the lateral slip ratio of a single steering wheel: ; Calculate the overall slip ratio: In the formula, For low-speed prevention, remove the zero threshold.

[0052] Secondly, the real-time output torque of each steering wheel drive motor and steering motor is collected, and the longitudinal and lateral forces at the wheel ends are calculated. Combined with the vertical load of a single wheel, the real-time adhesion utilization rate of the current steering wheel is obtained.

[0053] Among them, the actual adhesion utilization rate under the current slip ratio ,satisfy: In the formula, For the longitudinal force at the wheel end, The force is the lateral force at the wheel end. For a single-wheel vertical load, it can be distributed as a static axle load; among which, The formula for the longitudinal force at the wheel end is: In the formula, The motor output torque is determined based on current observation or a torque sensor. Where r is the reduction ratio and r is the wheel radius. To calibrate the rolling resistance torque. Lateral force. The calculation can be performed by observing and converting the steering motor torque, or by inversely calculating the acceleration using the vehicle dynamics equations and the IMU.

[0054] Then, a recursive least squares algorithm with a forgetting factor is used, with the initial value of kinematic slip ratio and real-time adhesion utilization rate as input, to identify the peak adhesion coefficient and optimal slip ratio parameters corresponding to the current road surface online, and update the constructed steering wheel slip-adhesion model.

[0055] The steering wheel slip-adhesion model is a model describing the relationship between the current slip ratio and the peak adhesion coefficient of the road surface, expressed as: In the formula, This represents the peak adhesion coefficient of the road surface. The optimal slip ratio (the slip ratio corresponding to peak adhesion).

[0056] The recursive least squares algorithm with a forgetting factor, in updating the constructed steering wheel slip-adhesion model, includes the following steps: considering that the slip ratio is a fast variable and the adhesion coefficient is a slow variable; calculating the initial value of the slip ratio according to the kinematic formula at the fast scale. The motor torque is calculated to determine the real-time wheel end force and obtain the current adhesion utilization rate. At a slow scale, by ; Deformation yields: Nonlinear recursive least squares is used, with updates performed at each step. Two parameters.

[0057] Based on the updated steering wheel slip-adhesion model, the dynamically corrected slip ratio is obtained through real-time adhesion utilization inversion, including: Calculated from real-time wheel end force It is the observable truth value, Substituting the inverse function of the adhesion model, we calculate the true slip ratio under dynamic constraints: Then, the initial value of the kinematic slip ratio is... With dynamic correction slip ratio Adaptive weighted fusion is performed to output the final real-time slip ratio of each steering wheel.

[0058] In a specific embodiment, to achieve more precise and stable positioning and attitude control, and improve the accuracy and efficiency of AGV parking positioning, considering factors such as the AGV's total pose error, real-time slip rate, and left-right slip asymmetry, fuzzy inference is used to upgrade from directly outputting scalar commands to first outputting vehicle-level motion parameters and then mapping them to scalar commands. This dynamically generates the single-step expected reversal angle, the maximum safety limit of each wheel's drive torque, and the pause observation time, better adapting to multi-wheel slip asymmetry conditions and ensuring that the target turning angle of each wheel satisfies the Ackerman cooperative kinematics constraints. Step S3 of the method also includes: First, the asymmetry of the AGV's left and right side slippage is calculated in real time, and the global maximum slippage rate is determined. Specifically, the asymmetry of the AGV's left and right side slippage... Global maximum slip ratio .

[0059] Secondly, using the current AGV's total pose error, global maximum slip ratio, and left and right side slip asymmetry as fuzzy inputs, the expected yaw rate ratio, expected translational speed ratio, and instantaneous steering center ICR offset factor are obtained through fuzzy inference. Based on the fuzzy inference output, the single-step expected reversing angle of the current control step, the maximum safety limit of each wheel drive torque, and the pause observation time after the reversing action are calculated.

[0060] Specifically, while keeping the single-step expected reversing angle, the maximum safety limit of each wheel drive torque, and the pause observation time after the reversing action unchanged, the three vehicle-level parameters of expected yaw rate ratio, expected translational speed ratio, and instantaneous steering center ICR offset factor are integrated. This not only directly reflects the balance between real-time slip risk and motion requirements, but also ensures that the output form remains unchanged through mapping, and naturally inherits the kinematic coordination capabilities carried by the vehicle-level parameters.

[0061] Among them, the three intermediate parameters at the vehicle body level are defined as the proportions of the desired yaw rate: , , Maximum angular velocity allowed for vehicle rotation; desired translational velocity ratio: , , The maximum linear velocity of the vehicle body center translation (resultant velocity); instantaneous steering center ICR offset factor. It is used to shift the instantaneous steering center laterally to adjust the load on the left and right wheels.

[0062] The mapping relationship from intermediate parameters to output variables, and the output mapping: The single-step desired reversal angle is obtained by multiplying the desired yaw rate ratio by the preset maximum single-step rotation angle: Δ ; To determine the maximum allowed turning angle in a single step, you can directly inherit the value. The adjustment results.

[0063] The maximum safety limit for the drive torque of each wheel is based on the smaller of the desired yaw rate ratio and the desired translational speed ratio, and a preset rated torque. And the reduction factor determined by the real-time slip risk; ; Depend on A comprehensive decision based on slip risk: Then, the torque limit is reduced by the reduction coefficient determined by the real-time slip risk to ensure that the upper limit of torque is amplified synchronously when the step length is large and the upper limit of torque is drastically reduced when there are multiple slips. Among them, the real-time slip risk is determined by quantifying the slip risk index and calculating the slip risk index value of each round according to the real-time slip rate, thereby obtaining the slip risk level and the reduction coefficient of the preset mapping of different slip risk levels.

[0064] The pause observation duration is set inversely based on the combined motion intensity (weighted root mean square) of the ratio of desired yaw rate to desired translational rate. The lower the motion intensity and the higher the risk of slippage, the longer the pause duration. Specifically, the settings are reversed based on the level of slippage risk and the intensity of the movement, such as: The lower the weighted root mean square value, the higher the risk of slippage and the longer the pause observation time.

[0065] The ICR offset factor, under multi-wheel slip asymmetric conditions, adjusts and corrects the instantaneous steering center position and the desired yaw center, so that the target steering angles of each wheel calculated based on the single-step desired reversal angle satisfy the Ackermann cooperative kinematic constraints. Specifically, in calculating the standard ICR position... At that time, offset correction is completed. , It is a preset maximum offset distance, which can be set to half the wheelbase or according to the vehicle geometry; That is, only lateral displacement. When At that time, the ICR shifts to the left (positive y-direction); At that time, it shifted to the right.

[0066] The rule design for fuzzy inference includes: the input is the total pose error of the AGV ( To positioning error , Composition Heading angle error The system includes global maximum slip ratio and left / right lateral slip asymmetry, with each input parameter having several levels, such as L and S. The outputs are the desired yaw rate ratio, desired translational speed ratio, and instantaneous steering center ICR offset factor. Each output parameter is mapped to a preset input parameter, such as: , , Furthermore, NB indicates a significant rightward shift of ICR and a significant leftward shift of PB. Some mapping settings are shown in Table 1 below: Table 1

[0067] In the table, "-" indicates that the variable is irrelevant or has any value. Based on the mapping table, output the values ​​of the expected yaw rate ratio, the expected translational rate ratio, and the instantaneous steering center ICR offset factor.

[0068] In addition, to further improve the ability to suppress slippage and the positioning accuracy during AGV positioning and attitude adjustment, a deep reinforcement learning network and preset fuzzy rules can be combined to adaptively select a control strategy based on the global slippage risk index, thereby more accurately controlling the single-step expected reversing angle, the maximum safety limit of the driving torque of each wheel, and the pause observation time after the reversing action; the method also includes: step S3 further includes: First, a deep reinforcement learning network is constructed. The construction steps include: defining the state as the total pose error of the AGV and the real-time slip rate; defining the action as a continuous action, controlling the expected single-step reversal angle, the maximum safe limit of the drive torque for each wheel, and the pause observation time after executing the reversal action; defining the reward function based on a weighted calculation of the arrival distance error reward and the slip penalty. Specifically, the arrival distance error reward can be positively rewarded based on a reduction in the distance error to the target position or reaching the positioning accuracy range, plus a preset value H; the slip penalty is calculated as follows: .

[0069] Secondly, the slip risk indicators for each round are quantified, and the slip risk indicator values ​​for each round are calculated based on the real-time slip rate. These values ​​are then combined and weighted to generate a global slip risk index. Specifically, the slip risk indicators for each round are as follows: In the process of generating a global slip risk index by combining the slip risk index values ​​of each wheel, the weight of each wheel can be determined based on the historical slip risk index frequency of each wheel. The higher the frequency, the greater the risk, and the higher the corresponding weight.

[0070] Then, the decision-making mode is selected based on the global slip risk index. When the global slip risk index is greater than the first threshold, it indicates severe slippage in multiple rounds. In this case, inference is performed based on preset fuzzy rules, mapping the intermediate parameters to output variables to output the expected single-step reversal angle, maximum safety limit, and pause observation duration. Pure fuzzy rules ensure extremely conservative output commands, preventing aggressive behavior. When the global slip risk index is less than the second threshold, it indicates that most wheels are healthy. A deep reinforcement learning network is used to calculate the expected single-step reversal angle, maximum safety limit, and pause observation duration, directly outputting commands to pursue attitude adjustment efficiency. When the global slip risk index is between the first and second thresholds, the output of the preset fuzzy rules and the output of the deep reinforcement learning network are weighted and fused. The fuzzy output provides a safety benchmark, and the RL output provides optimization corrections, ultimately calculating the final output.

[0071] In a specific embodiment, to improve the composite control law, the additional driving torque required to overcome the inertial torque of the entire vehicle is calculated and added to the total torque of each wheel. This pre-counters load changes caused by dynamic motion, allowing the AGV to better cope with the effects of dynamic motion during positioning and attitude adjustment, thereby improving the accuracy and stability of positioning and attitude adjustment. The method further includes: the composite control law in step S4 also includes: The inverse dynamic feedforward compensation term is based on the single-step expected commutation angular acceleration and the vehicle's rotational inertia. This compensation term is calculated based on the single-step expected commutation angular acceleration corresponding to the differential of the generated single-step expected commutation angular velocity, combined with the vehicle's rotational inertia about the vertical axis and the geometric layout parameters of the vehicle body and steering wheel. The additional driving torque required to overcome the vehicle's inertial torque is then added to the total torque of each wheel as a feedforward quantity to pre-counteract the load changes caused by dynamic motion.

[0072] Specifically, considering the planar motion of a four-wheeled independent steering wheel AGV, we establish the vehicle dynamics. Neglecting suspension and gradient, we set the vehicle mass as m and the yaw moment of inertia as... Each wheel in the vehicle coordinate system Steering angle The longitudinal and lateral forces of the wheel are Vehicle dynamics equations in the vehicle coordinate system: , , First, based on the generated single-step expected commutation angular velocity, the expected single-step commutation angular acceleration is obtained through differentiation or tracking differentiator.

[0073] Then, using inverse dynamics feedforward calculation: based on the known single-step desired commutation angular acceleration... Current status , And the lateral force that can be measured or estimated via a Kalman filter. (Through state estimation), the longitudinal forces of each wheel are obtained by weighted pseudo-inverse solution. This makes the above vehicle dynamics equations hold true. This is a system of linear equations: In the formula, , The lateral force vector that can be measured or estimated by a Kalman filter. ; Calculated based on the proportion of the desired translational velocity. A is obtained by differentiating the components; The matrix, with the i-th column being ; It is a generalized force generated by lateral force.

[0074] Finally, the longitudinal force of each wheel Multiply by the effective radius (the geometric layout parameters of the steering wheel) to obtain the additional driving torque required to overcome the inertial torque of the entire vehicle.

[0075] In a specific embodiment, to further improve the accuracy and stability of AGV positioning and attitude adjustment, the source of deviation during the positioning and attitude adjustment process is traced, and compensation parameters are generated and relevant parameters are corrected online in a targeted manner, effectively reducing positioning errors caused by inaccurate kinematic parameters, lag in dynamic response, or local slip events; the method further includes: the three-degree-of-freedom deviation recalibration calculation of pose during the pause in step S6 further includes: When generating the output of the current control step, the theoretical motion trajectory under ideal no-slip conditions is simultaneously calculated based on the AGV's kinematic model and the current initial pose, serving as a reference trajectory. Assuming ideal ground adhesion and pure rolling of each wheel (no longitudinal slippage, no lateral slippage), the forward integral calculates the vehicle's pose sequence throughout the entire motion time: In the formula, f is the differential equation of pure rolling motion. This step provides an ideal no-slip reference estimate, including the time history of position and attitude; the desired velocity spinor for a single step is calculated based on three vehicle-level intermediate parameters output by fuzzy inference. .

[0076] During the pause calibration phase, the actual motion trajectory is obtained by fusing multi-source sensor data through a Kalman filter. The actual motion trajectory is then dynamically compared with the reference trajectory to obtain the trajectory shape (including calculating the trajectory tangential direction deviation and curvature difference at each moment to identify offsets or random disturbances), convergence speed (comparing the time when key pose components (such as heading angle) reach the target value to determine whether the actual response is lagging or leading) and the difference in the final pose (the static deviation of the final pose).

[0077] Based on the differences, the sources of deviation are traced and distinguished, identifying whether the deviation originates from inaccurate kinematic parameters, dynamic response lag, or local slip events, and corresponding compensation parameters are generated. Kinematic parameter deviations generally manifest as whether the actual path bends or deviates, possibly due to inaccurate kinematic model parameters such as wheel wear or wheelbase calibration errors. Dynamic response lag generally manifests as a significantly slower convergence speed than the reference trajectory, or the occurrence of oscillations, potentially reflecting insufficient dynamic response or overshoot in the control system (such as a composite control law), requiring adjustments to the parameters of the composite control law. Local slip events generally manifest as a discrepancy between the actual endpoint error and the prediction based on slip rate estimation, potentially requiring verification and updating of the slip observation model and the observation equation parameters. The corresponding compensation parameters are then used to correct the AGV kinematic model parameters, composite control law parameters, and Kalman filter parameters online.

[0078] In a specific embodiment, to further improve the reliability and stability of the AGV positioning and attitude adjustment process and reduce positioning errors caused by slippage, the calculation method of the target steering angle and target wheel speed of each wheel is flexibly adjusted based on the comparison result of the global slippage risk index and the preset switching threshold, so as to more accurately cope with different slippage risk conditions. The method further includes: Step S5 further includes: When the global slip risk index is lower than the preset switching threshold, it is determined to be a low slip risk state. The wheels are working in the linear region, and the results of the inverse kinematics solution can be directly used as the target steering angle and target wheel speed of each wheel, that is, the wheel torque is not considered and it is purely kinematic allocation.

[0079] When the global slip risk index is equal to or higher than the preset switching threshold, it is determined to be a medium-to-high slip risk state. Torque allocation is introduced to actively utilize the adhesion potential of each wheel, further suppressing slip and improving tracking accuracy. The desired generalized force is allocated to each wheel tire force using a weighted pseudo-inverse or online optimization. Then, the steering angle and wheel speed / torque commands are obtained through the tire inverse model. The specific steps are as follows: First, the desired vehicle body velocity spinor is generated based on the single-step desired reversal angle and pose error. Specifically, based on the three vehicle-level intermediate parameters output by fuzzy inference, the output single-step desired velocity spinor is calculated and obtained. Determine the maximum permissible driving torque for each wheel. ; Global torque limit factor.

[0080] Secondly, the desired generalized force is generated through speed closed-loop control. Specifically, this is achieved through a speed closed-loop PID controller. Converted into the required resultant force and yaw moment: In the formula, , These are the estimated longitudinal / lateral velocities of the vehicle body. The estimated angular velocity of the vehicle body; the tire force vector to be assigned is defined. Efficiency matrix Determined by the current steering angle and wheel position. .

[0081] A weighted matrix is ​​constructed using the slip risk index values ​​of each round as weights. .

[0082] The desired generalized force is distributed into the desired longitudinal and lateral forces of each wheel using a weighted pseudo-inverse, and this distribution is constrained by the maximum safety limit of the driving torque of each wheel and the friction circle constraint. The formula for solving the desired tire force using the weighted pseudo-inverse is as follows: ; ; Friction circle constraint: , Corresponding coefficient; Torque limit: The final expected force is obtained by solving the problem. , .

[0083] Then, the final expected force is obtained based on the final solution. , and The target steering angle and target wheel speed of each wheel are calculated by using the tire inverse model.

[0084] like Figure 2 As shown in the figure, this application discloses a slip suppression control system for positioning and attitude adjustment of a four-wheeled independent steering wheel AGV, specifically including: The mode switching module 100 is used to monitor the AGV entering the preset fine positioning area and start the fine positioning and posture adjustment process, switching to the slip suppression control mode. The slip detection module 200 is used to collect the angular velocity of each steering wheel encoder, IMU data, and laser positioning data; it constructs a Kalman filter with vehicle coordinates, heading angle, translational velocity, and angular velocity as state variables, and iteratively calculates the state update equation and the observation equations of the fused encoder, IMU, and laser positioning to output the real vehicle motion state quantity without slip interference; based on the real vehicle motion state quantity, it calculates the slip ratio of each steering wheel to determine the total pose error of the AGV; The reversing control module 300 is used to take the current total pose error of the AGV and the real-time slip rate as input, and perform inference according to the preset fuzzy rules to generate the single-step expected reversing angle of the current control step, the maximum safety limit of the driving torque of each wheel, and the pause observation time after the reversing action is executed. The torque control module 400 is used to combine the output of the current control step with the slip ratio as the constraint input and to calculate the driving torque of each steering wheel using a composite control law. The composite control law includes: a PID closed-loop term based on real speed feedback, a moment of inertia feedforward term, a viscous and Coulomb friction compensation term, a ground load friction feedforward term, and a slip limit term that dynamically calculates the upper limit of torque based on the current slip ratio and a preset suppression coefficient. The motion coordination module 500 is used to solve the steering angle and wheel speed of the four wheels based on the current control step output and the principle of instantaneous rotation center, using inverse kinematics to ensure that the speed vectors of the four wheels converge at the same instantaneous rotation center. The suppression execution module 600 is used to perform single-step reversal based on the acquired drive torque, steering angle and wheel speed after each steering wheel is limited, and then enter the pause calibration stage. During the pause, the pose three-degree-of-freedom deviation recalibration calculation is performed; it is determined whether the pose deviation meets the preset fine positioning accuracy requirements; if not, it returns to the slip detection module to repeat the reversal control, torque control and cooperative motion until the requirements are met, and exits the attitude adjustment mode.

[0085] This application also discloses a computer-readable storage medium.

[0086] Specifically, the computer-readable storage medium stores a computer program that can be loaded by a processor and executed, such as the slip suppression control method for positioning and attitude adjustment of the four-wheel independent steering wheel AGV described above. The computer-readable storage medium includes, for example, various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0087] This application also discloses a computer device.

[0088] Specifically, the computer device includes a memory and a processor. The memory stores a computer program that can be loaded by the processor and executed to perform the slip suppression control method for positioning and attitude adjustment of the four-wheeled independent steering wheel AGV.

[0089] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Any feature disclosed in this specification (including the abstract and drawings) may be replaced by other equivalent or similar features unless specifically stated otherwise. That is, unless specifically stated otherwise, each feature is only one example of a series of equivalent or similar features.

Claims

1. A slip suppression control method for positioning and attitude adjustment of a four-wheeled independent steering wheel AGV, characterized in that, include: S1. Monitor the AGV to enter the preset fine positioning area and start the fine positioning and posture adjustment process, then switch to the slip suppression control mode. S2. Collect angular velocity data from the encoders of each steering wheel, IMU data, and laser positioning data; construct a Kalman filter with vehicle coordinates, heading angle, translational velocity, and angular velocity as state variables; iteratively calculate the state update equation and the observation equations of the fused encoder, IMU, and laser positioning to output the true vehicle motion state variables without slippage interference; calculate the slip ratio of each steering wheel based on the true vehicle motion state variables to determine the total pose error of the AGV; S3. Using the current total AGV pose error and real-time slip rate as input, reason according to the preset fuzzy rules to generate the single-step expected reversing angle of the current control step, the maximum safety limit of the driving torque of each wheel, and the pause observation time after the reversing action is executed. S4. Combining the current control step output with slip ratio as the constraint input, calculate the driving torque of each steering wheel using a composite control law; The composite control law includes: a PID closed-loop term based on real speed feedback, a moment of inertia feedforward term, a viscous and Coulomb friction compensation term, a ground load friction feedforward term, and a slip limit term that dynamically calculates the upper limit of torque based on the current slip ratio and a preset suppression coefficient. S5. Based on the current control step output, and using the principle of instantaneous rotation center, solve the steering angle and wheel speed of the four wheels using inverse kinematics to ensure that the speed vectors of the four wheels converge at the same instantaneous rotation center. S6. Based on the obtained driving torque, steering angle and wheel speed of each steering wheel after amplitude limiting, perform single-step reversal, and then enter the pause calibration stage. During the pause, perform pose three-degree-of-freedom deviation recalibration calculation. S7. Determine whether the pose deviation meets the preset precision positioning accuracy requirements; if not, return to step S3 to continue iteration; if it meets the requirements, exit the pose adjustment mode.

2. The slip suppression control method for positioning and attitude adjustment of a four-wheeled independent steering wheel AGV according to claim 1, characterized in that, Step S2 involves calculating the slip ratio of each steering wheel, including: The actual vehicle speed obtained from the filtering solution is converted into the longitudinal and lateral speeds in the coordinate system of each steering wheel center. Combined with the measured wheel speed and steering angle of each steering wheel, the initial value of the kinematic slip ratio of each steering wheel is calculated. The real-time output torque of each steering wheel drive motor and steering motor is collected, and the longitudinal and lateral forces at the wheel ends are calculated. Combined with the vertical load of a single wheel, the real-time adhesion utilization rate of the current steering wheel is calculated. A recursive least squares algorithm with a forgetting factor is used, with the initial value of the kinematic slip ratio and the real-time adhesion utilization rate as input, to identify the peak adhesion coefficient and the optimal slip ratio parameter corresponding to the current road surface online, and update the constructed steering wheel slip-adhesion model; the steering wheel slip-adhesion model is a model describing the relationship between the current slip ratio and the peak adhesion coefficient of the road surface. Based on the updated steering wheel slip-adhesion model, the dynamic correction slip ratio is obtained by inverting the real-time adhesion utilization rate. Then, the initial value of the kinematic slip ratio and the dynamic correction slip ratio are adaptively weighted and fused to output the final real-time slip ratio of each steering wheel.

3. The slip suppression control method for positioning and attitude adjustment of a four-wheeled independent steering wheel AGV according to claim 1, characterized in that, Step S3 also includes: Calculate the asymmetry of the AGV's left and right side slippage in real time and determine the global maximum slippage rate; Using the current AGV's total pose error, global maximum slip ratio, and left-right slip asymmetry as fuzzy inputs, fuzzy inference yields the desired yaw rate ratio, desired translational speed ratio, and instantaneous steering center ICR offset factor. Based on the fuzzy inference output, the expected single-step reversal angle, the maximum safety limit of each wheel drive torque, and the pause observation duration after the reversal action are calculated for the current control step. The expected single-step reversal angle is obtained by multiplying the expected yaw rate ratio by the preset maximum single-step rotation angle; the maximum safety limit of each wheel drive torque is based on... The determination is based on the smaller of the desired yaw rate ratio and the desired translational speed ratio, the preset rated torque, and a reduction coefficient determined by the real-time slip risk. The pause observation duration is set inversely based on the combined motion intensity of the desired yaw rate ratio and the desired translational speed ratio; the lower the motion intensity and the higher the slip risk, the longer the pause duration. The real-time slip risk is determined by quantifying the slip risk index and calculating the slip risk index value of each wheel based on the real-time slip rate. The ICR offset factor adjusts and corrects the instantaneous steering center position under multi-wheel slip asymmetry conditions. The fuzzy inference rule design includes: the inputs are the total pose error of the AGV, the global maximum slip ratio, and the left and right slip asymmetry, and each input parameter is set with several levels; the outputs are the expected yaw rate ratio, the expected translational rate ratio, and the instantaneous steering center ICR offset factor, and each output parameter is combined with the preset input parameters to form a mapping.

4. The slip suppression control method for positioning and attitude adjustment of a four-wheeled independent steering wheel AGV according to claim 3, characterized in that, Step S3 also includes: Construct a deep reinforcement learning network; the construction steps include: defining the state as the total pose error of the front AGV and the real-time slip rate; defining the action as a continuous action, controlling the single-step expected reversal angle, the maximum safety limit of the driving torque of each wheel, and the pause observation time after the reversal action; defining the reward function based on the weighted calculation of the arrival positioning distance error reward and the slip penalty; Quantify the slip risk indicators, calculate the slip risk indicator values ​​for each round based on the real-time slip rate, and combine the slip risk indicator values ​​for each round to generate a weighted global slip risk index; When the global slip risk index is greater than the first threshold, inference is performed based on the preset fuzzy rules, and the expected single-step reversal angle, maximum safety limit, and pause observation duration are output. When the global slip risk index is less than the second threshold, the expected single-step reversal angle, maximum safety limit, and pause observation duration are calculated using a deep reinforcement learning network. When the global slip risk index is between the first and second thresholds, the output of the preset fuzzy rules and the output of the deep reinforcement learning network are weighted and fused to obtain the final output.

5. The slip suppression control method for positioning and attitude adjustment of a four-wheeled independent steering wheel AGV according to claim 4, characterized in that, The composite control law in step S4 also includes: The inverse dynamic feedforward compensation term is based on the single-step expected commutation angular acceleration and the vehicle's rotational inertia. This compensation term is calculated based on the single-step expected commutation angular acceleration corresponding to the differential of the generated single-step expected commutation angular velocity, combined with the vehicle's rotational inertia about the vertical axis and the geometric layout parameters of the vehicle body and steering wheel. The additional driving torque required to overcome the vehicle's inertial torque is then added to the total torque of each wheel as a feedforward quantity to pre-counteract the load changes caused by dynamic motion.

6. The slip suppression control method for positioning and attitude adjustment of a four-wheeled independent steering wheel AGV according to claim 4, characterized in that, The pose three-degree-of-freedom deviation recalibration calculation performed during the pause in step S6 also includes: When generating the output of the current control step, the theoretical motion trajectory under ideal no-slip conditions is simultaneously calculated based on the kinematic model of the AGV and the current starting pose, and used as a reference trajectory. During the pause calibration phase, the actual motion trajectory is obtained by fusing multi-source sensor data through a Kalman filter, and the actual motion trajectory is dynamically compared with the reference trajectory to obtain the differences in trajectory shape, convergence speed and endpoint pose. Based on the differences, the sources of deviation are traced and distinguished, and the sources of deviation are distinguished as inaccurate kinematic parameters, dynamic response lag, or local slip events, and corresponding compensation parameters are generated. The corresponding compensation parameters are then used to correct the AGV kinematic model parameters, composite control law parameters, and Kalman filter parameters online.

7. A slip suppression control system for positioning and attitude adjustment of a four-wheeled independent steering wheel AGV, characterized in that, include: The mode switching module is used to monitor the AGV entering the preset fine positioning area and start the fine positioning and posture adjustment process, switching to the slip suppression control mode; The slip detection module is used to collect angular velocity data from the encoders, IMU data, and laser positioning data of each steering wheel; it constructs a Kalman filter with vehicle coordinates, heading angle, translational velocity, and angular velocity as state variables, and iteratively calculates the state update equation and the observation equations of the fused encoder, IMU, and laser positioning to output the true vehicle motion state variables without slip interference; based on the true vehicle motion state variables, it calculates the slip ratio of each steering wheel to determine the total pose error of the AGV; The reversing control module is used to take the current AGV total pose error and real-time slip rate as input, and perform inference according to preset fuzzy rules to generate the single-step expected reversing angle of the current control step, the maximum safety limit of the driving torque of each wheel, and the pause observation time after the reversing action is executed. The torque control module is used to calculate the driving torque of each steering wheel by combining the output of the current control step with the slip ratio as the constraint input and employing a composite control law. The composite control law includes: a PID closed-loop term based on real speed feedback, a moment of inertia feedforward term, a viscous and Coulomb friction compensation term, a ground load friction feedforward term, and a slip limit term that dynamically calculates the upper limit of torque based on the current slip ratio and a preset suppression coefficient. The motion coordination module is used to solve the steering angle and wheel speed of the four wheels based on the current control step output and the principle of instantaneous rotation center, using inverse kinematics to ensure that the speed vectors of the four wheels converge at the same instantaneous rotation center. The suppression execution module is used to perform single-step reversal based on the acquired drive torque, steering angle and wheel speed after each steering wheel is limited, and then enter the pause calibration stage. During the pause, the pose three-degree-of-freedom deviation recalibration calculation is performed; it is determined whether the pose deviation meets the preset fine positioning accuracy requirements; if not, it returns to the slip detection module to repeat the reversal control, torque control and cooperative motion until the requirements are met, and then exits the attitude adjustment mode.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the method as described in any one of claims 1 to 6.

9. A computer device, characterized in that, The computer device includes a memory, a processor, and a program stored in and executable on the memory, the program being executed by the processor to implement the steps of the method as described in any one of claims 1 to 6.

Images