Agricultural robot layered anti-skid control method

Abstract

The present application relates to a kind of agricultural robot layered antiskid control method, comprising the following steps: constructing vehicle-ground-actuator coupling controlled object model;Establish layered control architecture, obtain 4WD-4WS agricultural robot real-time running state and reference trajectory, and input layered control architecture;Using the predictive control strategy with disturbance feedforward compensation, calculate the expected generalized control force of whole vehicle;Using the optimal allocation strategy with constraint, the expected generalized control force of whole vehicle is distributed as the independent drive instruction and independent steering instruction of each wheel;The independent drive instruction and independent steering instruction of each wheel are issued to bottom layer actuator, and the real-time pose, motion state, wheel load and slip rate data of vehicle are collected by on-board sensor, feedback to upper path tracking control layer and lower drive force distribution layer, complete rolling horizon closed-loop control.Can significantly improve the operation accuracy and stability of agricultural robot in complex soft soil conditions.

Description

Technical Field

[0001] This invention relates to the field of automatic control technology for agricultural robots, and in particular to a layered anti-slip control method for agricultural robots. Background Technology

[0002] Due to terrain limitations, large agricultural machinery is difficult to operate in hilly and mountainous areas. Small, intelligent agricultural robots have become the core equipment for achieving mechanized and precise operations in these areas. The 4WD-4WS agricultural robot possesses high mobility with independent four-wheel drive and independent four-wheel steering. It can achieve various movement modes such as coordinated steering and crab-walking, making it well-suited for complex scenarios such as narrow ridges, field turns, and side slope operations in hilly and mountainous areas. It represents the core research and development direction for agricultural machinery equipment in hilly and mountainous regions.

[0003] However, current agricultural robot path tracking control technology still has significant technical shortcomings when dealing with harsh, unstructured environments in hilly and mountainous areas.

[0004] For example, existing patent CN112394734A discloses a vehicle trajectory tracking control method based on a linear model predictive control algorithm. This scheme mainly collects vehicle state information and reference trajectory, constructs a linear predictive model, performs rolling optimization, and outputs control commands. However, the predictive model of this scheme is entirely based on the assumption of a flat, rigid road surface, which is a quasi-static path tracking under ideal conditions. It does not involve an active observation and compensation mechanism for gravity components and soil perturbations under steep slope conditions in hilly terrain. When agricultural robots operate in steep or cross-slope environments, if they encounter lateral or longitudinal load transfer caused by gravity, the system cannot perform feedforward intervention, making it prone to uphill slippage or severe deviation on side slopes. In harsh off-road terrain, it is difficult to guarantee the trajectory accuracy of farmland operations.

[0005] For example, existing patent CN112224036A discloses a method and system for distributing torque to four wheels of a distributed drive electric vehicle. This method distributes the required torque to four independent wheels by constructing control targets such as yaw rate and center-of-gravity sideslip angle. However, this strategy only provides electrical and mechanical boundary constraints and has two major flaws: First, it lacks a dynamic anti-skid constraint mechanism based on the vehicle-soil contact mechanics. In hilly soft soil conditions, because the severe transfer of vertical load on the four wheels caused by longitudinal / lateral slope and acceleration / deceleration is not considered, simple motor torque limitation often cannot prevent the low-load side wheel from exceeding the effective adhesion limit of the soil. Second, in the optimization process of bottom-layer torque distribution, this scheme completely ignores the nonlinear physical processes of plastic settlement and shear failure in farmland soft soil. When the vehicle enters a muddy area, it is very easy for the wheels to slip excessively, leading to chassis settlement and causing the vehicle to completely lose its ability to get out of trouble and drive stably in complex farmland environments.

[0006] In summary, existing integrated or traditional hierarchical control architectures often face the contradiction between excessively high computing power requirements and insufficient real-time anti-slip performance. Summary of the Invention

[0007] In view of this, the purpose of this invention is to provide a layered anti-slip control method for agricultural robots, which can effectively improve the problems of poor path tracking accuracy, easy slippage and instability, and difficulty in embedding algorithms in agricultural machinery in hilly and mountainous areas.

[0008] The technical solution adopted by this invention to solve the above-mentioned technical problems is: a layered anti-slip control method for agricultural robots, comprising the following steps: S1. Based on the 4WD-4WS agricultural robot, construct a vehicle-ground-actuator coupled controlled object model. The vehicle-ground-actuator coupled controlled object model includes a vehicle dynamics model, a tire-soil contact mechanics model, and an actuator constraint model. S2. Establish a hierarchical control architecture, which consists of an upper path tracking control layer and a lower drive force distribution layer. Obtain the real-time operating status and reference trajectory of the 4WD-4WS agricultural robot and input them into the hierarchical control architecture. S3, Upper-level path tracking control: A predictive control strategy with disturbance feedforward compensation is adopted. Based on the real-time operating status and reference trajectory of the 4WD-4WS agricultural robot, the expected generalized control force of the whole vehicle is calculated. S4. Optimized distribution of lower driving force: An optimal distribution strategy with constraints is adopted. The limit of nonlinear dynamic adhesion force of a single wheel is calculated based on the tire-soil contact mechanics model and used as a hard constraint of dynamic anti-skid linear inequality. Combined with the actuator constraint model, the desired generalized control force of the whole vehicle is distributed into independent driving commands and independent steering commands for each wheel. S5. Closed-loop execution and status feedback: The independent drive commands and independent steering commands of each wheel are sent to the bottom actuator. The vehicle's real-time pose, motion state, wheel load and slip ratio data are collected by the on-board sensors and fed back to the upper path tracking control layer and the lower drive force distribution layer to complete the rolling time domain closed-loop control.

[0009] Furthermore, in step S1, the construction of the vehicle dynamics model includes: establishing a vehicle coordinate system with the robot's center of mass as the origin, introducing gravity components caused by longitudinal and lateral slopes, establishing a three-degree-of-freedom motion differential equation that includes longitudinal motion, lateral motion, and yaw motion, and integrating static geometric allocation, slope gravity projection, dynamic inertia transfer, and centrifugal force transfer to construct a four-wheel dynamic vertical load calculation model adapted to unstructured chemical conditions.

[0010] Furthermore, in step S1, the construction of the tire-soil contact mechanics model includes: based on the soil bearing settlement theory, calculating the dynamic settlement of the wheel and the bulldozing resistance by combining the real-time vertical load of a single wheel; calculating the maximum shear thrust of a single wheel based on the soil shear theory; combining the bulldozing resistance and the maximum shear thrust, deriving the real-time dynamic adhesion limit of a single wheel, and applying a safety margin correction and non-negative constraint to it. The coefficient of the safety margin correction is in the range of 0.7-0.9.

[0011] Furthermore, in step S2, the real-time operating status includes the 4WD-4WS agricultural robot's pose, speed, heading angle, wheel load, and slip rate; the control cycle of both layers of the hierarchical control architecture is 10ms, which is adapted to the real-time requirements of the embedded platform.

[0012] Furthermore, the safety margin correction coefficient is dynamically adjusted based on the real-time acquired longitudinal and lateral slope data and slip ratio data. When a preset slope threshold greater than 15° is detected, or the slip ratio fluctuation exceeds the preset safety range of 5%, the safety margin correction coefficient is set to a lower limit of 0.7.

[0013] Furthermore, in step S1, the actuator constraint model construction includes: establishing a first-order inertial dynamics model based on the 4WD-4WS agricultural robot drive system.

[0014] Furthermore, step S3 includes: constructing a nonlinear disturbance observer to observe the lumped disturbance caused by slope and environmental changes in real time, obtaining the real-time disturbance estimate, and substituting the real-time disturbance estimate as a feedforward compensation term into the LTV-MPC control algorithm for optimization, and outputting the desired generalized control force of the whole vehicle.

[0015] Furthermore, step S4 includes: combining the actuator constraint model to construct a quadratic optimization objective function that includes a generalized force distribution error penalty term and a drive energy consumption penalty term, wherein the weight of the drive energy consumption penalty term is inversely proportional to the single-wheel vertical load calculated in real time by the vehicle dynamics model; After linearizing the single-wheel nonlinear dynamic adhesion limit, a dynamic anti-skid linear inequality hard constraint is obtained. This constraint, along with the physical limit constraint and steering geometry constraint of the actuator constraint model, is used as an inequality constraint for quadratic programming to obtain the independent drive command and independent steering command for each wheel.

[0016] Furthermore, the load adaptive drive energy consumption penalty term in the quadratic optimization objective function, which targets the first... Weighting coefficient of each wheel The computational logic satisfies: in, The first one is solved in real time by the vehicle dynamics model Dynamic vertical load on each wheel The gain is adjusted to a preset constant.

[0017] Furthermore, the linearization process of the single-wheel nonlinear dynamic adhesion limit includes: calculating the single-wheel nonlinear dynamic adhesion limit, and performing a first-order Taylor expansion linearization process on the single-wheel nonlinear dynamic adhesion limit at the actual slip rate and dynamic load working point of the previous control cycle, transforming it into a dynamic anti-slip linear inequality hard constraint with the driving force increment as the independent variable.

[0018] The beneficial effects of this application are as follows: 1. This application is specifically designed for unstructured soft soil working conditions in hilly and mountainous areas, and constructs a vehicle dynamics model and a tire-soil contact mechanics model that deeply integrates slope disturbance and dynamic load transfer. This makes the control system fully conform to the actual operating characteristics of hilly farmland at the underlying physical logic level.

[0019] 2. This application linearizes the highly nonlinear single-wheel dynamic adhesion limit using a first-order Taylor expansion and directly applies it as a hard constraint condition for the dynamic anti-slip inequality in the underlying quadratic programming solution. This defensive anti-slip mechanism ensures that the driving command never exceeds the maximum shear limit of the soil, fundamentally eliminating the safety hazard of slippage and instability in hilly soft soil.

[0020] 3. The invention application adopts a two-layer decoupled architecture of upper-layer path tracking + lower-layer driving force allocation, which cleverly reduces the over-driving optimization problem involving complex nonlinear soil mechanics to a computationally efficient convex optimization problem. This architecture enables the upper and lower layers to maintain a synchronous control cycle of 10ms, perfectly paving the way for the engineering implementation of advanced anti-slip algorithms on low-computing-power vehicle platforms for agricultural robots.

[0021] 4. This invention introduces gravity components caused by longitudinal and lateral slopes into the vehicle dynamics model and constructs a nonlinear disturbance observer in the upper path tracking control layer. This allows for real-time observation of lumped disturbances caused by slope and environmental changes, obtaining real-time disturbance estimates. These real-time disturbance estimates are then used as feedforward compensation terms in the discrete linear time-varying prediction model to actively counteract environmental disturbances. This effectively suppresses problems such as uphill slippage and lateral deviation, greatly improving the trajectory tracking accuracy of agricultural robots in terrains with steep slopes, cross slopes, and soft soil. Attached Figure Description

[0022] Figure 1 This is a flowchart of the present invention.

[0023] Figure 2 This is a schematic diagram of the hierarchical control architecture in this invention.

[0024] Figure 3 This is a schematic diagram of the upper path tracking control layer in this invention.

[0025] Figure 4 This is a schematic diagram of the lower driving force distribution layer in this invention.

[0026] Figure 5 This is a diagram showing the 4WD-4WS agricultural robot of the present invention starting and maintaining a 20° slope.

[0027] Figure 6 This is a comparison chart of the tracking trajectory of the 4WD-4WS agricultural robot of this invention under the condition of double-line shifting on a 15° slope. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that, in the description of this invention, unless otherwise stated, "a plurality of" means two or more; the terms "upper," "lower," "left," "right," "inner," "outer," "front end," "rear end," "head," "tail," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0029] Please see Figure 1-6 This invention provides a layered anti-slip control method for agricultural robots, comprising the following steps: S1. Based on the 4WD-4WS agricultural robot, construct a vehicle-ground-actuator coupled controlled object model. The vehicle-ground-actuator coupled controlled object model includes a vehicle dynamics model, a tire-soil contact mechanics model, and an actuator constraint model.

[0030] The construction of the vehicle dynamics model includes: establishing a vehicle coordinate system with the robot's center of mass as the origin; introducing gravity components caused by longitudinal and lateral slopes; establishing a three-degree-of-freedom motion differential equation that includes longitudinal, lateral, and yaw motions; and integrating static geometric allocation, slope gravity projection, dynamic inertia transfer, and centrifugal force transfer to construct a four-wheel dynamic vertical load calculation model adapted to unstructured chemical conditions, and solving the vertical load of each wheel in real time. .

[0031] The construction of the tire-soil contact mechanics model includes: calculating the dynamic settlement of the wheel and the bulldozing resistance based on the soil bearing settlement theory and the real-time vertical load of a single wheel; calculating the maximum shear thrust of a single wheel based on the soil shear theory; deriving the real-time dynamic adhesion limit of a single wheel by combining the bulldozing resistance and the maximum shear thrust; and applying a safety margin correction and non-negative constraints to it. The coefficient of the safety margin correction ranges from 0.7 to 0.9. The coefficient of the safety margin correction is dynamically adjusted based on the real-time acquired longitudinal and lateral slope data and slip ratio data. When a preset slope threshold greater than 15° is detected, or a slip ratio fluctuation exceeds a preset safety range of 5%, the coefficient of the safety margin correction is set to a lower limit of 0.7.

[0032] The tire-soil contact mechanics model includes a model based on the Bekker-Janosi ground mechanics theory, targeting the cohesive soft soil environment of hilly farmland, and utilizing soil parameters (cohesion) obtained from the model. internal friction angle (etc.) Construct a tire-soil contact mechanics model. First, calculate the single-wheel bulldozing resistance based on the Bekker bearing model. Subsequently, the maximum shear thrust of a single wheel was calculated based on the Janosi-Hanamoto shear model. Finally, the real-time dynamic adhesion limit of a single wheel is derived. .

[0033] The actuator constraint model construction includes: establishing a first-order inertial dynamics model based on the 4WD-4WS agricultural robot drive system, integrating the physical torque limit and steering angular velocity limit of the motor to form physical hard constraint conditions.

[0034] S2. Establish a hierarchical control architecture, which consists of an upper DOB-MPC path tracking control layer and a lower QP drive force distribution layer. Obtain the real-time operating status and reference trajectory of the 4WD-4WS agricultural robot and input them into the hierarchical control architecture. The real-time operating status includes the 4WD-4WS agricultural robot's pose, speed, heading angle, wheel load, and slip rate. The control cycle of both layers of the hierarchical control architecture is 10ms, which is suitable for the real-time requirements of the embedded platform.

[0035] S3. Upper-level path tracking control: A predictive control strategy with disturbance feedforward compensation is adopted. Based on the real-time operating status and reference trajectory of the 4WD-4WS agricultural robot, the expected generalized control force of the whole vehicle is calculated.

[0036] Step S3 specifically includes: constructing a nonlinear disturbance observer (NDOB) to observe the lumped disturbance caused by slope and environmental changes in real time, obtaining the real-time disturbance estimate, and substituting the real-time disturbance estimate as a feedforward compensation term into the LTV-MPC control algorithm for optimization, outputting the desired generalized control force of the whole vehicle (desired longitudinal resultant force). Desired yaw moment ).

[0037] Specifically, the upper path tracking control layer also includes a tracking error settlement module. The tracking error settlement module calculates and outputs error state variables based on the received real-time running status and reference trajectory (reference pose, reference speed, path curvature). The error state variables and real-time disturbance estimates are substituted into the LTV-MPC control algorithm as feedforward compensation terms to output the vehicle's desired generalized control force.

[0038] S4. Optimized distribution of lower driving force: An optimal distribution strategy with constraints is adopted. The limit of nonlinear dynamic adhesion force of a single wheel is calculated based on the tire-soil contact mechanics model and used as a hard constraint of dynamic anti-skid linear inequality. Combined with the actuator constraint model, the desired generalized control force of the whole vehicle is distributed into independent driving commands and independent steering commands for each wheel.

[0039] Step S4 specifically includes: combining the actuator constraint model to construct a quadratic optimization objective function containing a generalized force distribution error penalty term and a drive energy consumption penalty term, wherein the weight of the drive energy consumption penalty term is inversely proportional to the single-wheel vertical load calculated in real time by the vehicle dynamics model. The load-adaptive drive energy consumption penalty term in the quadratic optimization objective function is specific to the first... Weighting coefficient of each wheel The computational logic satisfies: in, The first one is solved in real time by the vehicle dynamics model Dynamic vertical load on each wheel The gain is adjusted to a preset constant.

[0040] After linearizing the single-wheel nonlinear dynamic adhesion limit, a dynamic anti-slip linear inequality hard constraint is obtained. This constraint, along with the physical limit constraint and steering geometry constraint of the actuator constraint model, is used as an inequality constraint for quadratic programming solution. That is, it is solved in real time using the ActiveSetMethod to obtain the independent drive command and independent steering command of each wheel.

[0041] The linearization process of the single-wheel nonlinear dynamic adhesion limit includes: calculating the single-wheel nonlinear dynamic adhesion limit, and performing a first-order Taylor expansion linearization process on the single-wheel nonlinear dynamic adhesion limit at the actual slip rate and dynamic load working point of the previous control cycle, transforming it into a dynamic anti-slip linear inequality hard constraint with the driving force increment as the independent variable.

[0042] S5. Closed-loop execution and status feedback: The independent drive commands and independent steering commands of each wheel are sent to the bottom actuator. The vehicle's real-time pose, motion state, wheel load and slip ratio data are collected by the on-board sensors and fed back to the upper path tracking control layer and the lower drive force distribution layer to complete the rolling time domain closed-loop control.

[0043] This embodiment uses a 350kg 4WD-4WS agricultural robot specifically designed for hilly and mountainous terrain, including... Actuation mechanism: Four-wheel independent hub motor drive and four-wheel independent servo steering mechanism, supporting all modes of movement such as coordinated steering, crabbing, and stationary steering.

[0044] Vehicle-mounted sensors: GNSS+IMU integrated navigation system (real-time acquisition of attitude, speed, and heading angle), wheel speed encoder (real-time acquisition of wheel speed), tilt sensor (real-time acquisition of longitudinal / lateral slope angle). Wheel load sensors (real-time acquisition of vertical loads on all four wheels) ).

[0045] Controller: High-performance embedded chip STM32H7, meeting the requirements for real-time solving with low computing power. Control cycle: The upper path tracking control layer and the lower driving force distribution layer are synchronized at 10ms, adapting to the real-time requirements of embedded platforms.

[0046] The performance indicators verified in this embodiment under typical hilly and mountainous conditions are as follows: Figure 5 The graph shows the 4WD-4WS agricultural robot starting and maintaining a 20° slope. The transient backslip peak is ≤0.2m / s and converges quickly with no slippage throughout the process. The single-step solution time is stable at ≤6ms.

[0047] like Figure 6 The image shows a comparison of the tracking trajectories of an agricultural robot under double-tracking conditions on a 15° slope. The lateral tracking error of this application is also shown. 3cm, stable slip ratio 5%.

[0048] This invention also provides a layered anti-slip control system for agricultural robots, comprising a model building and updating module, an architecture and input module, an upper-level path tracking control module, a lower-level driving force optimization and allocation module, and a closed-loop execution and state feedback module, which are connected in sequence. The system is used to execute the methods described in any of the above-mentioned embodiments. The model building and updating module is used to build and update the vehicle dynamics model, the tire-soil contact mechanics model, and the actuator constraint model in real time. The architecture and input module is used to acquire the robot's real-time operating state and reference trajectory and input them into the layered control architecture. The upper-level path tracking control module is used to calculate the desired generalized control force of the entire vehicle using a predictive control strategy with disturbance feedforward compensation. The lower-level driving force optimization and allocation module is used to allocate the desired generalized control force of the entire vehicle into independent driving commands and independent steering commands for each wheel. The closed-loop execution and status feedback module is used to send the independent drive commands and independent steering commands to the underlying actuators; it collects real-time vehicle pose, motion state, wheel load and slip ratio data through on-board sensors and feeds them back to the upper-level path tracking control module and the lower-level drive force optimization allocation module to complete the rolling time domain closed-loop control; it also includes a reference trajectory generation module, which is used to generate reference pose, reference speed and path curvature.

[0049] The present invention also provides an agricultural robot, including a robot body, a four-wheel independent drive and four-wheel independent steering actuator, vehicle-mounted sensors, a controller GNSS and IMU integrated navigation system, a wheel speed encoder, an inclination sensor, and a wheel load sensor. The controller stores a computer program, and when the computer program is executed by a processor, it implements the steps of any of the above-described methods.

[0050] It should be noted that the above embodiments are only used to illustrate the present invention, but the present invention is not limited to the above embodiments. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A layered anti-slip control method for agricultural robots, characterized in that, Includes the following steps: S1. Based on the 4WD-4WS agricultural robot, construct a vehicle-ground-actuator coupled controlled object model. The vehicle-ground-actuator coupled controlled object model includes a vehicle dynamics model, a tire-soil contact mechanics model, and an actuator constraint model. S2. Establish a hierarchical control architecture, which consists of an upper path tracking control layer and a lower drive force distribution layer. Obtain the real-time operating status and reference trajectory of the 4WD-4WS agricultural robot and input them into the hierarchical control architecture. S3, Upper-level path tracking control: A predictive control strategy with disturbance feedforward compensation is adopted to calculate the desired generalized control force of the whole vehicle; S4. Optimized distribution of lower driving force: An optimal distribution strategy with constraints is adopted. The limit of nonlinear dynamic adhesion force of a single wheel is calculated based on the tire-soil contact mechanics model and used as a hard constraint of dynamic anti-skid linear inequality. Combined with the actuator constraint model, the desired generalized control force of the whole vehicle is distributed into independent driving commands and independent steering commands for each wheel. S5. Closed-loop execution and status feedback: The independent drive commands and independent steering commands of each wheel are sent to the bottom actuator. The vehicle's real-time pose, motion state, wheel load and slip ratio data are collected by the on-board sensors and fed back to the upper path tracking control layer and the lower drive force distribution layer to complete the rolling time domain closed-loop control.

2. The layered anti-slip control method for agricultural robots according to claim 1, characterized in that, In step S1, the construction of the vehicle dynamics model includes: establishing a vehicle coordinate system with the robot's center of mass as the origin, introducing gravity components caused by longitudinal and lateral slopes, establishing a three-degree-of-freedom motion differential equation that includes longitudinal motion, lateral motion and yaw motion, and integrating static geometric allocation, slope gravity projection, dynamic inertia transfer and centrifugal force transfer to construct a four-wheel dynamic vertical load calculation model adapted to unstructured chemical conditions.

3. The layered anti-slip control method for agricultural robots according to claim 2, characterized in that, In step S1, the construction of the tire-soil contact mechanics model includes: based on the soil bearing settlement theory, the dynamic settlement of the wheel and the bulldozing resistance are calculated by combining the real-time vertical load of a single wheel; based on the soil shear theory, the maximum shear thrust of a single wheel is calculated; combining the bulldozing resistance and the maximum shear thrust, the real-time dynamic adhesion limit of a single wheel is derived, and a safety margin correction and non-negative constraint are applied to it. The coefficient of the safety margin correction is in the range of 0.7-0.

9.

4. The layered anti-slip control method for agricultural robots according to claim 3, characterized in that, In step S2, the real-time operating status includes the 4WD-4WS agricultural robot's pose, speed, heading angle, wheel load, and slip rate; the control cycle of both layers of the hierarchical control architecture is 10ms, which is suitable for the real-time requirements of the embedded platform.

5. The layered anti-slip control method for agricultural robots according to claim 4, characterized in that, The safety margin correction coefficient is dynamically adjusted based on real-time acquired longitudinal and lateral slope data and slip ratio data. When a preset slope threshold greater than 15° is detected, or a slip ratio fluctuation exceeds a preset safety range of 5%, the safety margin correction coefficient is set to a lower limit of 0.

7.

6. The layered anti-slip control method for agricultural robots according to claim 5, characterized in that, In step S1, the actuator constraint model construction includes: establishing a first-order inertial dynamics model based on the 4WD-4WS agricultural robot drive system.

7. The layered anti-slip control method for agricultural robots according to claim 6, characterized in that, Step S3 includes: constructing a nonlinear disturbance observer to observe the lumped disturbance caused by slope and environmental changes in real time, obtaining the real-time disturbance estimate, and substituting the real-time disturbance estimate as a feedforward compensation term into the LTV-MPC control algorithm for optimization, and outputting the desired generalized control force of the whole vehicle.

8. The layered anti-slip control method for agricultural robots according to claim 7, characterized in that, Step S4 includes: Combining the actuator constraint model, a quadratic optimization objective function is constructed, which includes a generalized force distribution error penalty term and a drive energy consumption penalty term. The weight of the drive energy consumption penalty term is inversely proportional to the single-wheel vertical load calculated in real time by the vehicle dynamics model. After linearizing the single-wheel nonlinear dynamic adhesion limit, a dynamic anti-skid linear inequality hard constraint is obtained. This constraint, along with the physical limit constraint and steering geometry constraint of the actuator constraint model, is used as an inequality constraint for quadratic programming to obtain the independent drive command and independent steering command for each wheel.

9. The layered anti-slip control method for agricultural robots according to claim 8, characterized in that, The load adaptive drive energy consumption penalty term in the quadratic optimization objective function is for the first... Weighting coefficient of each wheel The computational logic satisfies: in, The first one is solved in real time by the vehicle dynamics model Dynamic vertical load on each wheel The gain is adjusted to a preset constant.

10. A layered anti-slip control method for agricultural robots according to claim 9, characterized in that, The linearization process of the single-wheel nonlinear dynamic adhesion limit includes: calculating the single-wheel nonlinear dynamic adhesion limit, and performing a first-order Taylor expansion linearization process on the single-wheel nonlinear dynamic adhesion limit at the actual slip rate and dynamic load working point of the previous control cycle, transforming it into a dynamic anti-slip linear inequality hard constraint with the driving force increment as the independent variable.

Images