A vehicle chassis system control method

Abstract

The application provides a vehicle chassis system control method, belonging to the technical field of vehicle control, comprising: upper layer control: taking the safety, comfort and stability of the whole vehicle as three optimization targets respectively, setting the weight of each optimization target according to the environmental information perceived by the vehicle-mounted sensor, realizing game balance solution to obtain the control target of each subsystem; middle layer control: taking the control target of each subsystem, environmental information and parameter information of the vehicle model as the input of machine learning, obtaining the value of the control variable of each subsystem after machine learning; lower layer control: transmitting the value of the control variable of each subsystem into model prediction, formulating the control decision according to the value of each control variable and the current state of the vehicle, realizing the control of the chassis system, and feeding back the control execution situation to the middle layer control. The application can effectively avoid the conflict between each system, and can guarantee the comfort, safety and stability of the vehicle.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of vehicle control, and particularly relates to a vehicle chassis system control method. BACKGROUND

[0002] With the rapid development of the automobile industry, more and more active control components and by-wire chassis components are integrated into the vehicle chassis system. However, with the increase in the number of actuators, the complexity of the system also increases. When the number of actuators exceeds the number of motion states that need to be controlled, an over-actuated system is formed. The over-actuated system allows multiple control modules to affect the same physical variable, which provides higher control flexibility, but also brings the risk of conflict between subsystems, which may damage the stability of the vehicle and threaten the safety of passengers. In the current chassis integrated control method, the decentralized architecture is simple to implement, but does not fully utilize shared information, while the centralized architecture provides globally optimal control performance, but lacks flexibility and fault safety redundancy. The multi-layer architecture attempts to balance these advantages and disadvantages, and optimizes the overall vehicle performance through hierarchical control methods. However, there are still problems such as high coordination complexity and limited real-time control capability, which still have deficiencies in enhancing the dynamic performance of the vehicle, as well as improving the driving safety, stability, comfort, etc. SUMMARY

[0003] The purpose of the present application is to provide a vehicle chassis system control method that can effectively avoid conflicts between systems while achieving integrated control of each subsystem of the chassis, and can ensure the comfort, safety and stability of the vehicle.

[0004] The present application is implemented by the following technical solutions:

[0005] A vehicle chassis system control method, the chassis system includes multiple subsystems, including upper layer control, middle layer control and lower layer control, specifically including the following steps:

[0006] Step S1, upper layer control: taking the safety, comfort and stability of the vehicle as three optimization objectives respectively, setting the weight of each optimization objective according to the environmental information perceived by the vehicle-mounted sensor, achieving game balance solution, and obtaining the control target of each subsystem;

[0007] Step S2, middle layer control: taking the control target of each subsystem, environmental information and parameter information of the vehicle model as the input of machine learning, obtaining the value of the control variable of each subsystem after machine learning;

[0008] Step S3, lower layer control: inputting the value of the control variable of each subsystem into model prediction, model prediction formulating control decisions according to the value of each control variable and the current state of the vehicle, realizing the control of the chassis system, and feeding back the control execution to the middle layer control.

[0009] Further, in the step S1, the control target of each subsystem is the direct effect after each subsystem executes the control command.

[0010] Further, in the step S1, the each subsystem includes a steering system, a braking system, a suspension system and a driving system, the control target of the steering system includes a steering angle and a steering acceleration, the braking system includes a braking torque and a braking acceleration, the suspension system includes a suspension stiffness and a suspension damping, and the driving system includes a driving torque and a driving acceleration.

[0011] Further, in the step S2, the machine learning includes a game layer and a machine learning network layer, the control target of each subsystem, the environmental information and the parameter information of the vehicle model are taken as the input of the game layer, the game layer outputs the control strategy of each subsystem, the control strategy of each subsystem and the feedback of the model predictive control execution are taken as a data set, the machine learning network layer analyzes and trains the data set to obtain the value of the control variable of each subsystem.

[0012] Further, in the step S3, the model prediction executes the control of the vehicle lateral dynamics, the longitudinal dynamics and the vertical dynamics according to the control decision.

[0013] Further, in the step S1, the safety is represented by the formula , the comfort is represented by the formula , and the stability is represented by the formula , wherein v x represents the minimum vehicle speed, d rel is a function of the relative distance of the front obstacle, f(d rel , v x ) is a distance and speed related collision risk assessment function, is the square of the body yaw rate, α1, α2, α3 are all coefficients, and the value range of each coefficient is [0, 1], is the longitudinal acceleration change rate, is the lateral acceleration change rate, RMS(a z ) is the root mean square value of the vertical vibration, β1, β2, β3 are all coefficients, and the value range of each coefficient is [0, 1], is the yaw rate error, is the roll angle change rate, and α is the tire side slip angle.

[0014] Further, in the step S1, the safety, the comfort and the stability of the whole vehicle are respectively taken as the three optimization targets, the weight of each optimization target is set, the utility function is established, and the constraint condition is set according to the environmental information sensed by the vehicle sensor, so as to realize the game balance solution.

[0015] The present application has the following beneficial effects:

[0016] 1、The present application firstly sets the safety, comfort and stability of the vehicle as three optimization targets in the upper layer control, sets the weight of each optimization target according to the environmental information perceived by the vehicle-mounted sensor, realizes the solution of game balance, obtains the control target of each subsystem, then sets the control target of each subsystem, environmental information and parameter information of the vehicle model as the input of machine learning in the middle layer control, obtains the value of the control variable of each subsystem after machine learning, finally sets the value of the control variable of each subsystem into the model prediction in the lower layer control, the model prediction executes the control command to realize the control of the chassis system, and the control execution is fed back to the middle layer control, so as to realize the combination of game theory, machine learning and model predictive control to form the control of the vehicle chassis system together, which can not only effectively coordinate each subsystem and reduce the conflict between the subsystems, but also can avoid the influence of the real-time control ability, improve the safety and comfort of passengers and increase the stability of the vehicle. BRIEF DESCRIPTION OF DRAWINGS

[0017] The present application will be further described in detail below with reference to the accompanying drawings.

[0018] Figure 1 The flowchart of the present application.

[0019] Figure 2 The principle diagram of the present application. DETAILED DESCRIPTION

[0020] As shown in Figure 1 and Figure 2 , the vehicle chassis system control method, the chassis system includes four subsystems, which are steering system, braking system, suspension system and driving system, the control target of each subsystem is the direct effect after the execution of the control command of each subsystem, that is, the control target of the steering system includes steering angle and steering acceleration, the braking system includes braking torque and braking acceleration, the suspension system includes suspension stiffness and suspension damping, and the driving system includes driving torque and driving acceleration.

[0021] The control method includes upper layer control, middle layer control and lower layer control, and specifically includes the following steps:

[0022] Step S1, upper layer control: setting the safety, comfort and stability of the vehicle as three optimization targets, setting the weight of each optimization target according to the environmental information perceived by the vehicle-mounted sensor, realizing the solution of game balance, and obtaining the control target of each subsystem;

[0023] Specifically, the safety is represented by the formula , the comfort is represented by the formula is expressed by formula wherein v x represents the minimum vehicle speed, d rel represents the minimum vehicle speed, d front = max(d front , 0) is a function of the relative distance of the front obstacle, d front is the actual distance from the current vehicle to the front vehicle, is a distance and speed related collision risk assessment function, and a' and β' are weight coefficients, is the square of the body yaw rate, and a1, a2, a3 are coefficients each having a range of [0, 1], is the longitudinal acceleration change rate, is the lateral acceleration change rate, and RMS(a z ) is the root mean square value of the vertical vibration, and β1, β2, β3 are coefficients each having a range of [0, 1], is the yaw rate error, is the roll angle change rate, and a is the tire side slip angle.

[0024] The weights of each optimization target are set and the utility function U = w s U1 + w c U2 + w t U3 is established, the constraint conditions including the side slip angle constraint |a|≤a max , the distance constraint d rel ≥ d min , the steering angle constraint and the driving force constraint |T|≤T max are set, and the game is used to solve to obtain each subsystem control target, wherein w s , w c , and w t are the weights of safety, comfort, and stability respectively, and w s + w c + w t = 1, the weights are determined according to the emphasis on safety, comfort, and stability in a specific environment based on the environmental information sensed by the vehicle-mounted sensor, x * is the decision variable, x i is the decision variable of the i-th control target, and x -i is the decision variable of other control targets, a max , d min , T max are the set threshold values, respectively.

[0025] The vehicle-mounted sensors include a laser radar installed on the top of the vehicle, the front bumper or below the window to detect obstacles, a radar installed on the bumper, the side body to measure the distance and speed, a GPS installed on the roof to locate, an inertial measurement unit installed in the central position inside the vehicle to measure the acceleration, angular velocity and direction of the vehicle, and a wheel speed sensor installed on the wheel to detect the rotation speed of the wheel.

[0026] In step S2, the middle layer control: the control target, environmental information and parameter information of the vehicle model of each subsystem are taken as the input of machine learning, and the value of the control variable of each subsystem is obtained after machine learning;

[0027] Specifically, the machine learning is a two-layer multi-modal machine learning architecture, which specifically includes a game layer and a machine learning network layer. The control target, environmental information and parameter information of the vehicle model of each subsystem are taken as the input of the game layer. The game layer outputs the control strategy of each subsystem through a convolution network. The control strategy of each subsystem and the feedback of the model predictive control execution are taken as a data set. The machine learning network layer analyzes and trains the data set to obtain the value of the control variable of each subsystem, realizes the warm start of the model predictive control, thereby reducing the iteration number of the model predictive control and speeding up the calculation time.

[0028] The vehicle model can be a 7-degree-of-freedom vehicle model, a 14-degree-of-freedom vehicle model or a higher-degree-of-freedom vehicle model.

[0029] The analysis and training of the data set include checking whether there are missing values, abnormal values or repeated values in the data set and processing them; processing the data features, such as normalizing the data features by using a convolutional neural network; dividing the data set into a training set, a validation set and a test set. According to the actual situation, a suitable neural network model is selected for training, and then verification and fine-tuning are performed. The value of the predicted control variable of each subsystem is obtained after the training is completed. The neural network model can be a convolutional neural network or a recurrent neural network.

[0030] In step S3, the lower layer control: the value of the control variable of each subsystem is transmitted into the model prediction. The model prediction formulates a control decision according to the value of each control variable and the current state of the vehicle, realizes the control of the chassis system, and feeds back the control execution to the middle layer control.

[0031] Specifically, after the machine learning network layer is trained, the model predictive control (MPC) is initialized, and then the model predictive control is started, and the performance detection and model updating are performed.

[0032] The model prediction executes the control of the vehicle lateral dynamics, longitudinal dynamics and vertical dynamics according to the control decision, i.e. the lateral dynamics include the steering system, the longitudinal dynamics include the suspension system, and the longitudinal dynamics include the brake system and the drive system.

[0033] The above merely provides the preferred embodiment of the present application, and therefore cannot limit the scope of the application. Any equivalent changes and modifications made in accordance with the patent scope and the content of the specification should still fall within the scope of the present application.

Claims

1. A vehicle chassis system control method, the chassis system comprising a plurality of subsystems, characterized by: The method comprises upper control, middle control and lower control, and specifically comprises the following steps: In step S1, the upper control, the safety, the comfort and the stability of the vehicle as a whole are taken as three optimization objectives, the weights of the optimization objectives are set according to the environmental information sensed by the vehicle-mounted sensor, the game balance is solved, and the control objectives of the subsystems are obtained. In step S2, the control objectives of the subsystems, the environmental information and the parameter information of the vehicle model are taken as the input of machine learning, and the values of the control variables of the subsystems are obtained after machine learning. In step S3, the values of the control variables of the subsystems are input into model prediction, the model prediction formulates the control decision according to the values of the control variables and the state of the vehicle, the control of the chassis system is realized, and the control execution is fed back to the middle control. The safety in the step S1 is represented by the formula The comfort is represented by the formula The stability is represented by the formula Wherein, Minimizes the vehicle speed, Is a function of the relative distance of the front obstacle, Is a distance and speed related collision risk assessment function, Is the square of the body yaw rate, , , All are coefficients, and the value range is [0, 1], Is the longitudinal acceleration change rate, Is the lateral acceleration change rate, Is the root mean square value of the vertical vibration, , , All are coefficients, and the value range is [0, 1], Is the yaw rate error, Is the roll angle change rate, Is the tire side slip angle.

2. The vehicle chassis system control method according to claim 1, characterized by: In step S1, the control objectives of the subsystems are the intuitive effects of the subsystems after executing the control commands.

3. The vehicle chassis system control method according to claim 2, characterized by: In step S1, the subsystems include the steering system, the braking system, the suspension system and the driving system, the control objectives of the steering system include the steering angle and the steering acceleration, the braking system includes the braking torque and the braking acceleration, the suspension system includes the suspension stiffness and the suspension damping, and the driving system includes the driving torque and the driving acceleration.

4. The vehicle chassis system control method according to claim 1 or 2 or 3, characterized by: In step S2, the machine learning includes the game layer and the machine learning network layer, the control objectives of the subsystems, the environmental information and the parameter information of the vehicle model are taken as the input of the game layer, the game layer outputs the control strategies of the subsystems, the control strategies of the subsystems and the feedback of the model predictive control execution are taken as the data set, the machine learning network layer analyzes and trains the data set, and the values of the control variables of the subsystems are obtained.

5. The vehicle chassis system control method according to claim 1 or 2 or 3, characterized by: In step S3, the model prediction executes the control of the vehicle lateral dynamics, the longitudinal dynamics and the vertical dynamics according to the control decision.

6. The vehicle chassis system control method according to claim 1 or 2 or 3, characterized by: In step S1, the safety, the comfort and the stability of the vehicle as a whole are taken as three optimization objectives, the weights of the optimization objectives are set according to the environmental information sensed by the vehicle-mounted sensor, the utility function is established, and the constraint conditions are set to realize the game balance solution.

Images