Off-road vehicle path tracking stability control method for complex terrain

Abstract

Disclosed in the invention is an off-road vehicle path tracking stability control method for a complex terrain. The method comprises the following steps: step 1, constructing a wheel-ground interaction model to obtain a longitudinal force Fx, a lateral force Fy and a vertical force Fz of the wheel-ground interaction; step 2, by means of the wheel-ground interaction model, constructing a side-tilting evaluation index J1 and a sideslip evaluation index J2, and carrying out weighted processing on the two indexes to obtain a comprehensive stability evaluation index RSS; step 3, training the comprehensive stability evaluation index, and adjusting the weight coefficient of J1 and J2 in the comprehensive stability evaluation index RSS; step 4, by means of the comprehensive stability evaluation index, constructing an MPC controller; step 5, by means of the MPC controller, carrying out stability control on a vehicle. In the invention, a comprehensive lateral stability evaluation index is provided, and the evaluation index is continuously optimized by means of a reinforcement learning algorithm, so that the evaluation index can more accurately represent the lateral stability.

Description

A path-following stability control method for off-road vehicles in complex terrain Technical Field

[0001] The present invention relates to the field of vehicle path tracking and stability control, and in particular to a path tracking stability control method for an off-road vehicle in complex terrain. Background Art

[0002] With the development of the times, the demand for forklifts in different scenarios has gradually increased, and the application environment has become increasingly complex. At the same time, due to the continuous improvement of computing power and the rise of demand, the concept of intelligence has gradually become popular. As a result, people's requirements for smart forklifts have gradually increased to ensure stability and tracking accuracy even in complex terrain.

[0003] Forklifts operating on complex terrain differ from those on structured roads due to the interaction between the wheels and the ground. Furthermore, forklifts operating on complex terrain are more susceptible to lateral instability and tracking errors. Lateral instability can be categorized as roll instability and sideslip instability, but these instability assessments often rely on a single metric, failing to comprehensively evaluate lateral stability. Furthermore, many intelligent vehicles only consider tracking accuracy, ignoring potential instability during the tracking process. If lateral instability occurs during tracking, the tracking process becomes meaningless.

[0004] Summary of the Invention

[0005] The object of the present invention is to provide a path tracking stability control method for an off-road vehicle in complex terrain, so as to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for controlling path tracking stability of an off-road vehicle in complex terrain comprises the following steps:

[0008] Step 1: Construct a wheel-ground interaction model to obtain the longitudinal force F of the wheel-ground interaction x , lateral force F y and vertical force F z ;

[0009] Step 2: Construct the roll evaluation index J1 and the sideslip evaluation index J2 through the wheel-ground interaction model, and perform weighted processing on the two indices to obtain the comprehensive stability evaluation index RSS;

[0010] Step 3: Train the comprehensive stability evaluation index and adjust the weight coefficients of J1 and J2 in the comprehensive stability evaluation index RSS;

[0011] Step 4: Construct an MPC controller through comprehensive stability evaluation indicators;

[0012] Step 5: Control the vehicle stability through the MPC controller.

[0013] As a further solution of the present invention, constructing a wheel-ground interaction model includes the following steps:

[0014] Step 1.1: Construct a static pressure-bearing model to obtain the static subsidence of the wheel;

[0015] Step 1.2: Calculate the wheel slip rate, construct a dynamic wheel sinking model, and obtain a normal stress calculation method that takes wheel sinking into account.

[0016] Step 1.3: Construct a shear model and obtain a shear stress calculation method that takes wheel sinking into account.

[0017] As a further solution of the present invention: the steps of establishing and optimizing the comprehensive stability evaluation index are as follows:

[0018] Step 2.1, establish the rollover evaluation index J1;

[0019] Step 2.2: Establish the side slip evaluation index R of the front and rear axles f and R r , select R f and R r The larger one is used as the vehicle sideslip evaluation index J2;

[0020] Step 2.3: Perform weighted processing on the rollover evaluation index J1 and the sideslip evaluation index J2 to obtain the comprehensive stability evaluation index

[0021] Step 2.4: Update the weight of RSS through the deep reinforcement learning algorithm TD3 algorithm.

[0022] As a further solution of the present invention: the steps of the comprehensive stability evaluation index reinforcement learning training method are as follows:

[0023] Step 3.1, define the state as s = [J1, J2];

[0024] Step 3.2, define the action as a=[dw 1, dw2], dw1 and dw2 are the weight coefficient adjustments of LTR and R respectively;

[0025] Step 3.3: Define the reward function for learning.

[0026] Step 3.4: Construct the policy network Actor and the evaluation network Critic. The Actor network consists of two networks: the Actor network and the Target Actor network. The Critic network consists of four networks: the Critic1 network, the Critic2 network, the Target Critic1 network, and the Target Critic2 network. The Actor network, the Critic1 network, and the Critic2 network are updated via gradients, while the Target Actor and Target Critic networks are updated via soft updates.

[0027] Step 4.5: Substitute the updated comprehensive stability evaluation index into the a priori valid stability control algorithm to serve as the basis for the stability control of the control algorithm. Use the state quantity s = [J1, J2] to determine whether the vehicle is unstable using the updated comprehensive stability evaluation index, and give a reward value based on the reward function.

[0028] As a further solution of the present invention: the critical value of the comprehensive stability evaluation index RSS is RSS th , when RSS>RSS th When the vehicle is judged to be laterally unstable, the comprehensive stability evaluation index RSS is limited to below the critical value during the path tracking process.

[0029] As a further solution of the present invention: the state variables of the MPC controller system The output variable is u=[δM z M f ] T , the output variable is

[0030] The objective function expression of the MPC controller system is as follows:

[0031] Among them, Q1, Q2, Q3 are the output state weight matrices, R is the control quantity weight matrix, ε is the relaxation factor, and its weight coefficient is ρ.

[0032] As a further solution of the present invention: the construction of the MPC controller system in step 4 includes the following steps:

[0033] Step 4.1: Construct a four-degree-of-freedom vehicle dynamics model system:

[0034] Step 4.2: Set the reference trajectory of path tracking to a sextic Bezier curve. The reference trajectory equation is:

[0035] Step 4.3: Determine the state variables of the MPC controller system The control variable is u=[δM z M f ] T , the output variable is The vehicle system is represented as follows:

[0036] Step 4.4: Discretize the vehicle system:

[0037] Where A(k) = 1 + TA(t), B(k) = TB(t), I is the identity matrix, and T is the sampling time.

[0038] Step 4.5: Use the control increment Δu(k|t)=u(k|t)-u(k-1|t) to replace the control variable. The system can be expressed as:

[0039] Where,

[0040] Step 4.6: Express the predicted output as follows

[0041] In the formula

[0042] ΔU(k)=[Δu(k∣t) Δu(k+1∣t)…Δu(k+N c ∣t)] T

[0043] Among them, N p is the prediction time domain, N c is the control time domain, and satisfies N c <N p ;

[0044] Step 4.7: The constructed objective function expression is as follows:

[0045] Among them, Q1, Q2, Q3 are the output state weight matrices, R is the control quantity weight matrix, ε is the relaxation factor, and its weight coefficient is ρ.

[0046] As a further solution of the present invention: Step 5 comprises the following steps:

[0047] Step 5.1, input vehicle dynamic parameters into the vehicle desired path MPC path tracking controller;

[0048] Step 5.2, the MPC path tracking controller calculates the vehicle's control variable parameters by inputting the vehicle dynamic parameters and the vehicle's desired path;

[0049] Step 5.3: The vehicle receives the control variable parameters and controls the vehicle through the control variable parameters.

[0050] As a further solution of the present invention: the control variable parameters include control variables such as front wheel angle, roll moment and yaw moment, and the control variables such as front wheel angle, roll moment and yaw moment are respectively achieved by controlling the steering wheel angle, the support force of the suspension and the longitudinal force of the four wheels.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] 1. The present invention takes into account roll instability and sideslip instability in complex terrain, establishes roll instability evaluation indicators and sideslip instability evaluation indicators, combines the two, and proposes a comprehensive lateral stability evaluation index. The evaluation index is continuously optimized through a reinforcement learning algorithm, so that the evaluation index can more accurately represent lateral stability.

[0053] 2. The present invention designs an MPC controller, which comprehensively considers the accuracy and stability of path tracking, ensuring that the forklift can improve the accuracy of trajectory tracking in complex terrain without lateral instability. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] FIG1 is a control flow chart of this embodiment;

[0055] FIG2 is a flowchart of comprehensive stability evaluation index optimization in this embodiment. DETAILED DESCRIPTION

[0056] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0057] Referring to FIG1 , an embodiment of the present invention shows a stability control method for an off-road forklift during path tracking in complex terrain. The control steps are as follows:

[0058] Step 1: First, build the wheel-ground interaction model to obtain the longitudinal force F of the wheel-ground interaction x , lateral force F y and vertical force F z .

[0059] The steps to establish the wheel-ground interaction model are as follows:

[0060] Step 1.1: Construct a static pressure-bearing model to obtain the static subsidence of the wheel;

[0061] Step 1.2: Calculate the wheel slip rate, build a dynamic wheel sinking model, and obtain a normal stress calculation method that takes wheel sinking into account.

[0062] Where s is the slip rate, N is the dynamic sinking index, which is related to the slip rate, θ1 is the wheel approach angle, θ2 is the wheel departure angle, and θ m is the maximum stress angle;

[0063] Step 1.3: Construct a shear model and obtain a shear stress calculation method that takes wheel sinking into account. The shear stress can be decomposed into τ along the x-axis of the wheel coordinate system. x and τ in the y direction y ;

[0064] Therefore, the resultant force F in the horizontal direction can be obtained x and the vertical force F z :

[0065] When the vehicle turns, the wheels generate two lateral forces, F u The y-axis component of the shear stress τ y (θ) causes, F s is produced by lateral bulldozing, namely:

[0066] Where:

[0067] Step 2: Based on the forces obtained from the wheel-ground interaction model in step 1, construct the vehicle's roll evaluation index J1 and sideslip evaluation index J2, and perform weighted processing on the two indices to obtain the comprehensive stability evaluation index RSS (see Figure 2).

[0068] The steps for establishing and optimizing the comprehensive stability evaluation index are as follows:

[0069] Step 2.1: Establish rollover evaluation index J1

[0070] Step 2.2: Establish the side slip evaluation index R of the front and rear axles f and R r

[0071] Select R f and R r The larger one is used as the vehicle sideslip evaluation index J2 J2=max(R f ,R r )

[0072] Step 2.3: Perform weighted processing on the rollover evaluation index J1 and the sideslip evaluation index J2 to obtain the comprehensive stability evaluation index

[0073] Step 2.4: Update the weight of RSS through the deep reinforcement learning algorithm TD3 algorithm.

[0074] Step 3: Train the comprehensive stability evaluation index through reinforcement learning and adjust the weight coefficients of J1 and J2 in the comprehensive stability evaluation index RSS.

[0075] The steps of strengthening learning training method for comprehensive stability evaluation index are as follows:

[0076] Step 3.1, define the state as s = [J1, J2]

[0077] Step 3.2: Define the action as a = [Δw1, Δw2], where Δw1 and Δw2 are the weight coefficient adjustments of J1 and J2 respectively.

[0078] Step 3.3. Define the learned reward function R t : R t =k1R1+k2R2

[0079] Where R1 is the roll evaluation reward value, R2 is the sideslip evaluation reward value, as shown in the following formula, and k1 and k2 are the corresponding coefficients of the reward values.

[0080] Step 3.4: Construct a policy network (Actor) and a criticism network (Critic). The Actor network consists of two networks: the Actor network and the Target Actor network. The Critic network consists of four networks: the Critic1 network, the Critic2 network, the Target Critic1 network, and the Target Critic2 network. The Actor network inputs the state and outputs the action. The Critic network inputs the state and action and outputs the Q value.

[0081] The update process of the Critic network parameters is as follows:

[0082] Step 3.4.1. Target Actor Network Prediction t+1 Action a in statet+1 ;

[0083] Step 3.4.2, a t+1 The action adds noise ε as the input of the Target Critic1 network and the Target Critic2 network. The purpose of adding noise is to increase the diversity of exploration by adding noise to the output action of the Target Actor network, thereby making the Q value of the next step more accurate.

[0084] Step 3.4.3. Compare the Q values ​​output by the two Target Critic networks and select the smaller Q value to get the actual value y:

[0085] Step 3.4.4, the actual value y and the Q1(s output by the Critic network t ,a t ) and Q2(s t ,a t ) respectively and calculate the Loss function:

[0086] Step 3.4.5: Use the gradient descent method to update the parameters ω1 and ω2 of the critic network:

[0087] The Actor network uses Q1 (or Q2) calculated by the Cirtic1 (or Critic2) network as the Loss function and updates the parameter θ by gradient ascent:

[0088] The Target Actor and Target Critic networks are updated through soft updates: ω' i ←τω i +(1-τ)ω' i θ'←τθ+(1-τ)θ'

[0089] Step 3.5: Substitute the updated comprehensive stability evaluation index into the effective prior stability control algorithm and use it as the basis for the stability control of the control algorithm. Use the state quantity s = [J1, J2] to determine whether the vehicle is unstable when controlled by the stability control actuator using the updated comprehensive stability evaluation index, and give a reward value based on the reward function.

[0090] Step 4: Construct an MPC controller that takes into account the path tracking accuracy and stability during the tracking process.

[0091] Step 4.1: Construct a four-degree-of-freedom vehicle dynamics model system:

[0092] Step 4.2: Set the reference trajectory of path tracking to a sextic Bezier curve. The reference trajectory equation is:

[0093] Step 4.3: Determine the state variables of the MPC controller system The control variable is u=[δM z M f ] T , the output variable is The vehicle system is represented as follows:

[0094] Step 4.4: Discretize the vehicle system:

[0095] Where A(k)=1+TA(t), B(k)=TB(t), I is the unit matrix, and T is the sampling time.

[0096] Step 4.5: Use the control increment Δu(k|t)=u(k|t)-u(k-1|t) to replace the control variable. The system can be expressed as:

[0097] Where,

[0098] Step 4.6: Express the predicted output as follows

[0099] In the formula

[0100] ΔU(k)=[Δu(k∣t)Δu(k+1∣t)…Δu(k+N c ∣t)] T

[0101] Among them, N p is the prediction time domain, N c is the control time domain, and satisfies N c <N p .

[0102] Step 4.7: The constructed objective function expression is as follows:

[0103] Among them, Q1, Q2, Q3 are the output state weight matrices, R is the control quantity weight matrix, ε is the relaxation factor, and its weight coefficient is ρ.

[0104] Step 4.8: Constrain the control quantity, control quantity variable and output quantity, ucmax min Δucmax min ycmax min

[0105] Step 5: Distribute the front wheel angle, suspension force, and longitudinal force of the four wheels according to the output calculated by the MPC controller.

[0106] Step 5.1, input vehicle dynamic parameters into the vehicle desired path MPC path tracking controller;

[0107] Step 5.2, the MPC path tracking controller calculates the vehicle's control variable parameters by inputting the vehicle dynamic parameters and the vehicle's desired path;

[0108] Step 5.3: The vehicle receives the control variable parameters and controls the vehicle through the control variable parameters.

[0109] The comprehensive stability evaluation index RSS is greater than the critical value RSS th When the vehicle is considered to be laterally unstable, the instability is lateral instability or rollover and side sliding. During the path tracking process, the comprehensive stability evaluation index RSS is limited to below the critical value.

[0110] This method is used to control the stability of an off-road forklift during path tracking in complex terrain. Vehicle dynamics parameters are input into an MPC path-following controller. The MPC control objective function is designed to consider lateral position and lateral stability, along with dynamic constraints, control increments, and output constraints. This approach then calculates appropriate control variable values. The control variables, front wheel angle, roll moment, and yaw moment, are achieved by controlling the steering wheel angle, suspension support force, and four-wheel longitudinal force, respectively.

[0111] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. A path tracking stability control method for an off-road vehicle in complex terrain, characterized in that: The following steps are involved: Step 1: Construct a wheel-ground interaction model to obtain the longitudinal force F of the wheel-ground interaction x , lateral force F y and vertical force F z ; Step 2: construct the roll evaluation index J1 and the sideslip evaluation index J2 through the wheel-ground interaction model, and perform weighted processing on the two indices to obtain the comprehensive stability evaluation index RSS; Step 3, training the comprehensive stability evaluation index, and adjusting the weight coefficients of J1 and J2 in the comprehensive stability evaluation index RSS; Step 4: Construct an MPC controller through comprehensive stability evaluation indicators; Step 5: Perform vehicle stability control through the MPC controller.

2. A path tracking stability control method for an off-road vehicle in complex terrain according to claim 1, characterized in that: Building a wheel-ground interaction model involves the following steps: Step 1.1, construct a static pressure-bearing model to obtain the static settlement of the wheel; Step 1.2, calculate the slip rate of the wheel, construct a dynamic sinking model of the wheel, and obtain a normal stress calculation method considering wheel sinking. Step 1.3: Construct a shear model and obtain a shear stress calculation method that takes wheel sinking into consideration.

3. The method for controlling the path tracking stability of an off-road vehicle in complex terrain according to claim 1, characterized in that: The steps for establishing and optimizing the comprehensive stability evaluation index are as follows: Step 2.1, establish the rollover evaluation index J1; Step 2.2: Establish the side slip evaluation index R of the front and rear axles f and R r , select R f and R r The larger one is used as the vehicle sideslip evaluation index J2; Step 2.3: Perform weighted processing on the rollover evaluation index J1 and the sideslip evaluation index J2 to obtain the comprehensive stability evaluation index Step 2.4: Update and train the weight of RSS using the deep reinforcement learning algorithm TD3.

4. The method for controlling the path tracking stability of an off-road vehicle in complex terrain according to claim 3, characterized in that: The steps of the comprehensive stability evaluation index reinforcement learning training method are as follows: Step 3.1, define the state as s = [J1, J2]; Step 3.2, define the action as a=[dw1,dw2], where dw1 and dw2 are the weight coefficient adjustments of LTR and R respectively; Step 3.3: Define the reward function for learning. Step 3.4, construct the policy network Actor and the evaluation network Critic, the Actor network includes two networks: Actor network and Target Actor network, the Critic network includes four networks: Critic1 network, Critic2 network, Target Critic1 network, Target Critic2 network; the Actor network, Critic1 network and Critic2 network are updated by gradient, and the Target Actor and Target Critic networks are updated by soft update; Step 4.5: Bring the updated comprehensive stability evaluation index into the effective prior stability control algorithm each time, and use it as the basis for the stability control of the control algorithm. Use the state quantity s = [J1, J2] to determine whether the vehicle is unstable using the updated comprehensive stability evaluation index, and give a reward value based on the reward function.

5. The method for controlling the path tracking stability of an off-road vehicle in complex terrain according to claim 3, characterized in that: The critical value of the comprehensive stability evaluation index RSS is RSS th , when RSS>RSS th When the vehicle is judged to be laterally unstable, the comprehensive stability evaluation index RSS is limited to below the critical value during the path tracking process.

6. The method for controlling the path tracking stability of an off-road vehicle in complex terrain according to claim 1, characterized in that: The state variables of the MPC controller system The output variable is u=[δ M z M f ] T , the output variable is The objective function expression of the MPC controller system is as follows: Among them, Q1, Q2, Q3 are the output state weight matrices, R is the control quantity weight matrix, ε is the relaxation factor, and its weight coefficient is ρ.

7. A method for controlling the path tracking stability of an off-road vehicle in complex terrain according to claim 6, characterized in that: The construction of the MPC controller system in step 4 includes the following steps: Step 4.1: Construct a four-degree-of-freedom vehicle dynamics model system: Step 4.2: Set the reference trajectory of path tracking to a sixth-order Bezier curve. The reference trajectory equation is: Step 4.3: Determine the state variables of the MPC controller system The control variable is u=[δ M z M f ] T , the output variable is The vehicle system is represented as follows: Step 4.4: Discretize the vehicle system: Where A(k)=1+TA(t), B(k)=TB(t), I is the unit matrix, and T is the sampling time; Step 4.5: Use the control increment Δu(k|t)=u(k|t)-u(k-1|t) to replace the control variable. The system can be expressed as: In the formula, Step 4.6: Express the predicted output as follows In the formula Among them, N p is the prediction time domain, N c is the control time domain, and satisfies N c <N p ; Step 4.7: The constructed objective function expression is as follows: Among them, Q1, Q2, Q3 are the output state weight matrices, R is the control quantity weight matrix, ε is the relaxation factor, and its weight coefficient is ρ.

8. The method for controlling the path tracking stability of an off-road vehicle in complex terrain according to claim 1, characterized in that: The step 5 comprises the following steps: Step 5.1, input vehicle dynamic parameters into the vehicle desired path MPC path tracking controller; Step 5.2, the MPC path tracking controller calculates the control variable parameters of the vehicle by inputting the vehicle dynamic parameters and the vehicle desired path; Step 5.3: The vehicle receives the control variable parameters and controls the vehicle through the control variable parameters.

9. A method for controlling the path tracking stability of an off-road vehicle in complex terrain according to claim 8, characterized in that: The control variable parameters include control variables front wheel angle, roll moment and yaw moment, and the control variables front wheel angle, roll moment and yaw moment are respectively achieved by controlling the steering wheel angle, the support force of the suspension and the longitudinal force of the four wheels.

Images