A Cyber-Physical System for Electro-hydraulic Controller-by-Wire Chassis and Its Game Theory Optimization Method

Abstract

This invention discloses a cyber-physical system for an electro-hydraulic drive-by-wire chassis and its game-theoretic optimization method, comprising: an electro-hydraulic drive-by-wire chassis big data unit, a vehicle-road-cloud traffic information unit, an electro-hydraulic drive-by-wire chassis domain control unit, a vehicle condition unit, a road condition unit, and an execution unit. The electro-hydraulic drive-by-wire chassis big data unit communicates with the vehicle-road-cloud traffic information unit and the electro-hydraulic drive-by-wire chassis domain control unit. The vehicle-road-cloud traffic information unit also communicates with the vehicle condition unit and the road condition unit. The electro-hydraulic drive-by-wire chassis domain control unit also communicates with the execution unit. This invention improves vehicle energy efficiency and stability while maintaining driving comfort by optimizing the control of the electro-hydraulic drive-by-wire chassis domain control unit within a cyber-physical system environment. It not only handles the coupling relationships between various chassis subsystems but also achieves optimal performance under different operating conditions by coordinating steering and drive control.

Description

Technical Field

[0001] This invention belongs to the field of control optimization for drive-by-wire chassis, specifically relating to a cyber-physical system for electro-hydraulic drive-by-wire chassis and its game-theoretic optimization method. Background Technology

[0002] With the development of intelligent connected vehicle technology and new energy vehicles, electro-hydraulic drive-by-wire chassis systems have begun to be applied in autonomous vehicles. Traditional chassis systems, due to limitations in their mechanical structure, have certain limitations in control precision and response speed. Electro-hydraulic drive-by-wire chassis, by eliminating mechanical connections and employing electronic components such as sensors, electronic control units, and motors, achieve higher degrees of control freedom and response speed, thereby improving vehicle handling performance and safety.

[0003] Optimization of drive-by-wire chassis systems primarily involves integrating the chassis system and adding a game theory system to enhance controllability and comfort. Chinese invention patent application CN202110225619.2, entitled "A Drive-by-Wire Chassis Cyber-Physical System and Control Method in Intelligent Transportation Environment," proposes integrating hub motors with the suspension, eliminating traditional structures such as engines and clutches, simplifying the chassis structure, and improving active safety and operational stability. However, this patent neglects personalized user needs, resulting in poor control effectiveness for different driving styles. Chinese invention patent application CN202110471884.9, entitled "Intelligent Connected Vehicle Adaptive Obstacle Avoidance System Based on Drive-by-Wire Steering and Game Theory Results," proposes an intelligent connected vehicle adaptive obstacle avoidance system based on drive-by-wire steering and game theory results. It reports the collected information about the driver, vehicle, and environment during the game to a central server, improving obstacle avoidance capabilities after training. However, it lacks an evaluation of the dynamic changes in the weight factors of the evaluation network, resulting in poor generalization and fault tolerance.

[0004] Compared to vehicles equipped with traditional chassis, electro-hydraulic drive-by-wire chassis systems offer advantages in handling performance and safety. However, when different subsystems of the electro-hydraulic drive-by-wire chassis attempt to achieve the same control objective, conflicts may arise, affecting the overall performance of the chassis. Therefore, integrating different levels of traffic information physical systems and electro-hydraulic drive-by-wire chassis in real-world scenarios, and considering factors such as vehicle stability constraints and energy consumption to perform game-theoretic optimization of the electro-hydraulic drive-by-wire chassis, is of great significance in fully leveraging its control performance. Summary of the Invention

[0005] To address the shortcomings of the existing technologies, the present invention aims to provide a cyber-physical system for electro-hydraulic drive-by-wire chassis and its game-theoretic optimization method. By optimizing the control of the electro-hydraulic drive-by-wire chassis domain control unit within a cyber-physical system environment, the energy efficiency and stability of the vehicle are improved, while also ensuring driving comfort. The method of this invention not only handles the coupling relationships between various chassis subsystems but also achieves optimal performance under different operating conditions by coordinating steering and drive control.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] The present invention discloses a cyber-physical system for an electro-hydraulic drive-by-wire chassis, comprising: an electro-hydraulic drive-by-wire chassis big data unit, a vehicle-road-cloud traffic information unit, an electro-hydraulic drive-by-wire chassis domain control unit, a vehicle condition unit, a road condition unit, and an execution unit; the electro-hydraulic drive-by-wire chassis big data unit communicates with the vehicle-road-cloud traffic information unit and the electro-hydraulic drive-by-wire chassis domain control unit respectively; the vehicle-road-cloud traffic information unit also communicates with the vehicle condition unit and the road condition unit; the electro-hydraulic drive-by-wire chassis domain control unit also communicates with the execution unit.

[0008] The vehicle condition unit includes: a vehicle condition storage module and a vehicle condition sensor group module;

[0009] The vehicle condition sensor module is used to collect vehicle status data and send the collected data to the vehicle status storage module.

[0010] The vehicle status storage module is used to receive and store the data sent by the vehicle status sensor group module, and send the data to the first information storage module in the vehicle-road-cloud traffic information unit;

[0011] The road condition unit includes: a road condition storage module and a road condition sensor group module;

[0012] The road condition sensor module is used to collect road condition data and send the collected data to the road condition storage module.

[0013] The road condition storage module is used to receive and store the data sent by the road condition sensor group module, and send the data to the first information storage module in the vehicle-road-cloud traffic information unit;

[0014] The vehicle-road-cloud traffic information unit includes: a first information storage module and a game optimization module;

[0015] The first information storage module transmits data and stores data with the electro-hydraulic drive-by-wire chassis big data unit, vehicle condition unit, road condition unit, and game optimization module, respectively.

[0016] The game optimization module is used to optimize the active front wheel steering and direct yaw torque based on the data in the first information storage module. The optimization result is the actual output of the active front wheel steering and the actual output of the direct yaw torque. The optimization result is sent to the electro-hydraulic drive-by-wire chassis big data unit.

[0017] The electro-hydraulic drive-by-wire chassis big data unit includes: a data management module, a machine learning module, and an evaluation network module;

[0018] The data management module transmits data and stores data with the first information storage module, the electro-hydraulic drive-by-wire chassis domain control unit, the evaluation network module, and the machine learning module.

[0019] The evaluation network module is used to estimate the co-state of the evaluation network for the data collected by the vehicle condition sensor group module and the road condition sensor group module, calculate the target co-state of the data, and obtain the error function of the evaluation network by subtracting the estimated co-state of the evaluation network from the target co-state. The gradient descent algorithm is used to optimize the error function of the evaluation network, and the optimization result is sent to the data management module and the first information storage module.

[0020] The machine learning module is used to judge the driving conditions based on the decision tree model learned by the ideal operating database through the data collected by the vehicle condition sensor group module and the road condition sensor group module, optimize the model at the same time, and transmit the vehicle driving condition data to the data management module and the first information storage module.

[0021] The electro-hydraulic drive-by-wire chassis domain control unit includes: a second information storage module and an execution network module;

[0022] The second information storage module transmits data and stores data with the data management module, the execution unit, and the execution network module, respectively.

[0023] The execution network module is used to generate control commands using a BP neural network based on the actual output of active front wheel steering and the actual output of direct yaw torque data in the second information storage module, and then send the control commands to the second information storage module.

[0024] The execution unit includes: a drive-by-wire chassis steering module and a hub motor drive module;

[0025] The drive-by-wire chassis steering module is used to perform active front wheel steering based on control commands generated by the execution network module;

[0026] The hub motor drive module is used to execute direct yaw torque according to the control commands generated by the execution network module.

[0027] Furthermore, the decision tree model is used for classification and regression tasks; it reads data from the ideal operating database; according to the Gini index method, it selects a feature as the splitting criterion for the node, and splits the data in the ideal operating database into subsets based on the selected feature; it repeats the Gini index method for each subset, selects the best feature for splitting, and continues to split the data until the stopping condition is met; for new vehicle state data and road state data, it starts from the root node and follows the comparison results of feature values ​​until the leaf node is reached, and the value of the leaf node is the driving condition result.

[0028] Furthermore, the Gini index method measures the impurity of the dataset; the smaller the Gini index, the purer the dataset, and the feature with the smallest Gini index is selected for splitting.

[0029] Furthermore, the stopping conditions include: reaching a preset maximum tree depth, each leaf node containing fewer than a preset minimum number of samples, and the dataset no longer having significant splits.

[0030] Furthermore, the ideal operation database is an offline synchronous database containing a large amount of historical data under different driving conditions. It stores the optimal operation and control parameters of the vehicle under different driving conditions, which are derived from ideal operation data collected during experiments, simulations, and actual driving.

[0031] Furthermore, the driving conditions include: urban driving conditions, rural road driving conditions, special road driving conditions, and mixed driving conditions.

[0032] Furthermore, the data collected by the vehicle condition sensor module includes: vehicle slip ratio, center of gravity sideslip angle, longitudinal vehicle speed, yaw rate, longitudinal acceleration, and lateral acceleration.

[0033] Furthermore, the data collected by the road condition sensor module includes: the number and location of other vehicles around the vehicle, the ground adhesion coefficient, and the road slope.

[0034] Furthermore, the drive-by-wire chassis steering module includes: a steering motor controller, a steering actuator motor, a rack and pinion gear, and wheels; the steering motor controller controls the steering actuator motor through control commands, and the steering actuator motor is connected to the wheels through a rack and pinion gear; the steering motor controller receives control commands sent by the execution network module, drives the steering actuator motor to generate torque, and the torque is transmitted to the wheels under the action of the rack and pinion gear to realize the vehicle steering operation and complete the active front wheel steering.

[0035] Furthermore, the hub motor drive module includes: a drive motor controller and a hub motor; the drive motor controller controls the hub motor through control commands, and the hub motor is mechanically connected to the wheel; the drive motor controller receives control commands generated by the execution network module, drives the hub motor to generate driving force, and the hub motor generates yaw torque to complete the direct yaw torque.

[0036] This invention also provides a game-theoretic optimization method for a cyber-physical system of an electro-hydraulic drive-by-wire chassis. Based on the above system, the steps are as follows:

[0037] 1) Collect vehicle status data and road status data;

[0038] 2) Estimate the co-state of the evaluation network for the data collected in step 1), calculate the target co-state of the data, calculate the difference between the estimated co-state of the evaluation network and the target co-state to obtain the error function of the evaluation network, and optimize the error function to obtain the optimization result.

[0039] 3) Determine the driving conditions of the data collected in step 1) to obtain vehicle driving condition data;

[0040] 4) Using the data collected in step 1), the optimization results in step 2), and the vehicle driving condition data in step 3), game optimization is performed to obtain the actual output of active front wheel steering and the actual output of direct yaw torque.

[0041] 5) Calculate the control commands for active front wheel steering and direct yaw moment using the actual output of active front wheel steering and the actual output of direct yaw moment obtained in step 4).

[0042] 6) Complete vehicle control according to the control instructions obtained in step 5).

[0043] Furthermore, the estimation of the co-state of the network in step 2) is expressed as follows:

[0044]

[0045] In the formula, For X(k),ω AFS Estimation of the costate equation for the (l+1)th iteration of active front wheel steering under the given state. For X(k),ω DYC Estimate the costate equation of the direct yaw moment in the (l+1)th iteration under the given condition. Let ω be the augmented state variable at time k. AFS ω represents the weight coefficients of the intermediate layer of the active front wheel steering neural network. DYC ω represents the weight coefficients of the intermediate layer in a direct yaw moment neural network. AFS(l+1)ω is the weighting coefficient for the (l+1)th iteration of the direct yaw moment. DYC(l+1) φ is the weighting coefficient for the (l+1)th iteration of the direct yaw moment. AFS(l+1) Let φ be the basis function for the (l+1)th iteration of the active front wheel steering. DYC(l+1) It is the basis function for the (l+1)th iteration of the direct yaw moment.

[0046] Furthermore, the expression for the target co-state in step 2) is as follows:

[0047]

[0048] In the formula, D AFS D is the weight matrix for active front wheel steering. DYC Q is the weight matrix for the direct yaw moment. AFS Q represents the tracking error of active front wheel steering. DYC For the tracking error of the direct yaw moment, λ AFS(l+1) (X(k)) is the co-state equation for the (l+1)th iteration of the active front wheel steering in state X(k), λ DYC(l+1) (X(k)) is the co-state equation for the (l+1)th iteration of the direct yaw moment in state X(k), γ AFS γ is a discount factor for the active front steering performance index. DYC This is a discount factor for the direct yaw moment performance index.

[0049] Furthermore, the expression for the error function of the network evaluation in step 2) is:

[0050]

[0051] In the formula, e λAFS(l+1) (k) is the error function of the (l+1)th iteration of the active front wheel steering at time k, e λDYC(l+1) (k) is the error function of the (l+1)th iteration of the direct yaw moment at time k.

[0052] Furthermore, the specific steps of step 4) are as follows:

[0053] 41) Using the data collected in step 1), the optimization results in step 2), and the vehicle driving condition data in step 3), calculate the state output equation of the closed-loop Nash equilibrium and the state equation of the reference signal system.

[0054] 42) Calculate the tracking error e of active front wheel steering. AFS (k) and the tracking error of the direct yaw moment e DYC (k);

[0055] 43) Calculate the cost function to enable the vehicle to track the reference trajectory while minimizing energy consumption;

[0056] 44) Calculate the state-space equation and performance index function of the augmented game system;

[0057] 45) Calculate the game output of the weighting factors.

[0058] Furthermore, the state output equation of the closed-loop Nash equilibrium in step 41) is:

[0059]

[0060] In the formula, x(k+1) is the state variable at time k+1, x(k) is the state variable at time k, and u AFS (k) represents the input for active front wheel steering at time k, u DYC (k) represents the input quantity of the direct yaw moment at time k, A is the coefficient matrix of the state variables, B AFS B is the coefficient matrix for the active front wheel steering input. DYC C is the coefficient matrix for the direct yaw moment input. AFS C is the output matrix for active front wheel steering. DYC The output matrix for the direct yaw moment, y AFS (k) represents the actual output of the active front wheel steering at time k, y DYC (k) represents the actual output of the direct yaw moment at time k;

[0061] The state equation of the reference signal system is:

[0062]

[0063] In the formula, The state of the reference signal system at time k+1, Let r be the state of the reference signal system at time k. AFS (k) represents the desired output of the active front wheel steering reference signal system at time k, r DYC (k) represents the desired output of the direct yaw moment reference signal system at time k, F is the system matrix corresponding to the reference signal system, and G... AFS G is the output matrix of the active front wheel steering reference signal system system matrix. DYC This is the output matrix of the system matrix for the direct yaw moment reference signal system.

[0064] Furthermore, in step 42), the tracking error of active front wheel steering and direct yaw moment is defined as follows:

[0065]

[0066] In the formula, e AFS(k) represents the tracking error of the active front wheel steering at time k, e DYC (k) represents the tracking error of the direct yaw moment at time k.

[0067] Furthermore, the cost function in step 43) is:

[0068]

[0069] in, for The cost function of active front wheel steering under certain conditions. for The cost function of the direct yaw moment under the condition, r AFS (e AFS (i),u AFS (i)) is e AFS (i),u AFS (i) The desired output of the active front wheel steering reference signal system under the given state, r DYC (e DYC (i),u DYC (i)) is e DYC (i),u DYC (i) The desired output of the direct yaw moment reference signal system at time k under state (i), r AFS (e AFS (i),u AFS (i)) and r DYC (e DYC (i),u DYC (i)) is represented as follows:

[0070]

[0071] In the formula, R AFS R is the control cost weight matrix for active front wheel steering. DYC This is the control cost weight matrix for the direct yaw moment.

[0072] Furthermore, the state-space equation and performance index function of the augmented game system in step 44) are as follows:

[0073]

[0074]

[0075] In the formula, Let be the augmented state variables at time k+1. Let k be the augmented state variable at time k. This is the coefficient matrix of the augmented state variables. This is the coefficient matrix for the augmented active front wheel steering input. J is the coefficient matrix of the augmented direct yaw moment input. AFS (X(k)) is the cost function of active front wheel steering in state X(k), J DYC (X(k)) is the cost function of the direct yaw moment in the state of X(k).

[0076] Furthermore, in step 45), based on dynamic non-cooperative game theory, the closed-loop Nash strategy... It holds if and only if the following recursive relation exists:

[0077]

[0078] in:

[0079]

[0080] In the formula, J at time k AFS (X(k)) intermediate variable, J at time k DYC (X(k)) intermediate variable, Let k be the optimal value for active front wheel steering. This represents the optimal value of the direct yaw moment at time k.

[0081] Furthermore, step 5) specifically includes:

[0082] 51) Based on the actual output of the active front wheel steering and the actual output of the direct yaw moment in step 4), calculate the input estimate of the active front wheel steering in the lth iteration and the input estimate of the direct yaw moment in the lth iteration.

[0083] 52) Calculate the target value for the l-th iteration of the execution network;

[0084] 53) The error function of the execution network is obtained by taking the difference between the input estimate of the l-th iteration and the target value of the execution network;

[0085] 54) Adopting the adaptive rule of gradient descent, the weights of the execution network are updated, and the optimal control commands for active front wheel steering and direct yaw moment are generated. The vehicle is then controlled according to the control commands.

[0086] Further, in step 51), based on the actual output of the active front wheel steering and the actual output of the direct yaw moment output in step 4), the input estimates for the l-th iteration of the active front wheel steering and the input estimates for the l-th iteration of the direct yaw moment are calculated:

[0087]

[0088] in, For X(k),θ AFS The input estimation for the l-th iteration of active front wheel steering under the given conditions. For X(k),θ DYC The input estimate of the direct yaw moment in the l-th iteration under the given state, θ AFS θ represents the weight coefficients of the intermediate layer of the active front wheel steering neural network. DYC ψ represents the weight coefficients of the intermediate layer in a direct yaw moment neural network. AFS(l) (X(k)) is the basis function for the l-th iteration of active front wheel steering in state X(k), ψ DYC(l) (X(k)) is the basis function of the l-th iteration of the direct yaw moment in the state of X(k).

[0089] Further, the target value for the l-th iteration of the network in step 52) is expressed by the following formula:

[0090]

[0091] Among them, u AFS(l) (X(k)) is the input quantity for the l-th iteration of active front wheel steering in state X(k); u DYC(l) (X(k)) is the input quantity of the direct yaw moment in the l-th iteration under the state of X(k); λ AFS(l) (X(k+1)) is the co-state equation for the l-th iteration of active front wheel steering in state X(k+1), λ DYC(l) (X(k+1)) is the co-state equation for the l-th iteration of the direct yaw moment in the state X(k+1).

[0092] Further, in step 53), based on the target value of the execution network, the error function of the execution network is obtained as follows:

[0093]

[0094] Among them, e uAFS(l) (k) represents the target value of the active front wheel steering network in the l-th iteration, e uDYC(l) (k) is the target value of the network for the l-th iteration of the direct yaw moment.

[0095] The beneficial effects of this invention are:

[0096] This invention utilizes the execution network module in the electro-hydraulic steerable chassis domain control unit to read the actual output of the active front wheel steering and the actual output of the direct yaw moment from the second information storage module. Then, it calculates the optimal control commands for the steerable chassis module and the wheel hub motor drive module using a BP neural network structure. The neural network possesses strong fitting capabilities, allowing it to comprehensively consider various complex constraints, thereby enabling the vehicle to have a suitable control scheme.

[0097] This invention solves the problem of potential conflicts and a prisoner's dilemma that may occur when the electro-hydraulic drive-by-wire chassis control unit completes multiple control objectives by using a game optimization module in the vehicle-road-cloud traffic information unit, thus ensuring the overall performance of the chassis.

[0098] This invention utilizes the evaluation network module within the big data unit of the electro-hydraulic control-by-wire chassis, and employs the learning rule of the steepest descent method to dynamically send feedback information to the game optimization module. This enables the game optimization module to possess better generalization and fault tolerance capabilities, solving the curse of dimensionality problem in dynamic programming. It provides a practical theory and method for the optimal control problem of high-dimensional complex nonlinear systems of electro-hydraulic control-by-wire chassis. Attached Figure Description

[0099] Figure 1 This is a schematic diagram of the system of the present invention.

[0100] Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0101] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to embodiments and accompanying drawings. The content mentioned in the embodiments is not intended to limit the present invention.

[0102] Reference Figure 1 As shown, the present invention provides a cyber-physical system for an electro-hydraulic drive-by-wire chassis, comprising: an electro-hydraulic drive-by-wire chassis big data unit, a vehicle-road-cloud traffic information unit, an electro-hydraulic drive-by-wire chassis domain control unit, a vehicle condition unit, a road condition unit, and an execution unit; the electro-hydraulic drive-by-wire chassis big data unit communicates with the vehicle-road-cloud traffic information unit and the electro-hydraulic drive-by-wire chassis domain control unit respectively; the vehicle-road-cloud traffic information unit also communicates with the vehicle condition unit and the road condition unit; the electro-hydraulic drive-by-wire chassis domain control unit also communicates with the execution unit.

[0103] The vehicle condition unit includes: a vehicle condition storage module and a vehicle condition sensor group module;

[0104] The vehicle condition sensor module is used to collect vehicle status data and send the collected data to the vehicle status storage module.

[0105] The vehicle status storage module is used to receive and store the data sent by the vehicle status sensor group module, and send the data to the first information storage module in the vehicle-road-cloud traffic information unit;

[0106] The road condition unit includes: a road condition storage module and a road condition sensor group module;

[0107] The road condition sensor module is used to collect road condition data and send the collected data to the road condition storage module.

[0108] The road condition storage module is used to receive and store the data sent by the road condition sensor group module, and send the data to the first information storage module in the vehicle-road-cloud traffic information unit;

[0109] The vehicle-road-cloud traffic information unit includes: a first information storage module and a game optimization module;

[0110] The first information storage module transmits data and stores data with the electro-hydraulic drive-by-wire chassis big data unit, vehicle condition unit, road condition unit, and game optimization module, respectively.

[0111] The game optimization module is used to optimize the active front wheel steering and direct yaw torque based on the data in the first information storage module. The optimization result is the actual output of the active front wheel steering and the actual output of the direct yaw torque. The optimization result is sent to the electro-hydraulic drive-by-wire chassis big data unit.

[0112] The electro-hydraulic drive-by-wire chassis big data unit includes: a data management module, a machine learning module, and an evaluation network module;

[0113] The data management module transmits data and stores data with the first information storage module, the electro-hydraulic drive-by-wire chassis domain control unit, the evaluation network module, and the machine learning module.

[0114] The evaluation network module is used to estimate the co-state of the evaluation network for the data collected by the vehicle condition sensor group module and the road condition sensor group module, calculate the target co-state of the data, and obtain the error function of the evaluation network by subtracting the estimated co-state of the evaluation network from the target co-state. The gradient descent algorithm is used to optimize the error function of the evaluation network, and the optimization result is sent to the data management module and the first information storage module.

[0115] The machine learning module is used to judge the driving conditions based on the data collected by the vehicle condition sensor group module and the road condition sensor group module through the decision tree model learned by the ideal operation database, optimize the model at the same time, and transmit the vehicle driving condition data to the data management module and the first information storage module.

[0116] The decision tree model is used for classification and regression tasks. It reads data from the ideal operating database; selects a feature as the splitting criterion for a node using the Gini index method; splits the data in the ideal operating database into subsets based on the selected feature; repeats the Gini index method for each subset, selecting the best feature for splitting, and continues splitting the data until a stopping condition is met; for new vehicle state data and road state data, it starts from the root node and proceeds according to the comparison results of feature values ​​until a leaf node is reached, where the value of the leaf node is the driving condition result.

[0117] The Gini index method measures the impurity of a dataset; the smaller the Gini index, the purer the dataset, and the feature with the smallest Gini index is selected for splitting.

[0118] The stopping conditions include: reaching the preset maximum tree depth, each leaf node containing fewer than the preset minimum number of samples, and the dataset no longer having significant splits.

[0119] The ideal operation database is an offline synchronous database containing a large amount of historical data under different driving conditions. It stores the optimal operation and control parameters of the vehicle under different driving conditions, which are derived from ideal operation data collected during experiments, simulations and actual driving.

[0120] The electro-hydraulic drive-by-wire chassis domain control unit includes: a second information storage module and an execution network module;

[0121] The second information storage module transmits data and stores data with the data management module, the execution unit, and the execution network module, respectively.

[0122] The execution network module is used to generate control commands using a BP neural network based on the actual output of active front wheel steering and the actual output of direct yaw torque data in the second information storage module, and then send the control commands to the second information storage module.

[0123] The execution unit includes: a drive-by-wire chassis steering module and a hub motor drive module;

[0124] The drive-by-wire chassis steering module is used to perform active front wheel steering based on control commands generated by the execution network module;

[0125] The hub motor drive module is used to execute direct yaw torque according to the control commands generated by the execution network module.

[0126] Specifically, the driving conditions include: urban driving conditions, rural road driving conditions, special road driving conditions, and mixed driving conditions.

[0127] Specifically, the data collected by the vehicle condition sensor module includes: vehicle slip ratio, center of gravity sideslip angle, longitudinal vehicle speed, yaw rate, longitudinal acceleration rate, and lateral acceleration.

[0128] Specifically, the data collected by the road condition sensor module includes: the number and location of other vehicles around the vehicle, the ground adhesion coefficient, and the road slope.

[0129] Specifically, the drive-by-wire chassis steering module includes: a steering motor controller, a steering actuator motor, a gear rack, and wheels; the steering motor controller controls the steering actuator motor through control commands, and the steering actuator motor is connected to the wheels through a gear rack; the steering motor controller receives control commands sent by the execution network module, drives the steering actuator motor to generate torque, and the torque is transmitted to the wheels under the action of the gear rack to realize the vehicle steering operation and complete the active front wheel steering.

[0130] Specifically, the hub motor drive module includes: a drive motor controller and a hub motor; the drive motor controller controls the hub motor through control commands, and the hub motor is mechanically connected to the wheel; the drive motor controller receives control commands generated by the execution network module, drives the hub motor to generate driving force, and the hub motor generates yaw torque to complete the direct yaw torque.

[0131] Reference Figure 2 As shown, the present invention also provides a game-theoretic optimization method for a cyber-physical system for an electro-hydraulic drive-by-wire chassis. Based on the above system, the steps are as follows:

[0132] 1) Collect vehicle status data and road status data;

[0133] Specifically, road condition data includes: urban driving conditions, rural road driving conditions, special road driving conditions, and mixed driving conditions.

[0134] Specifically, vehicle status data includes: vehicle slip ratio, center of gravity sideslip angle, longitudinal vehicle speed, yaw rate, longitudinal acceleration, and lateral acceleration.

[0135] 2) Estimate the co-state of the evaluation network for the data collected in step 1), calculate the target co-state of the data, calculate the difference between the estimated co-state of the evaluation network and the target co-state to obtain the error function of the evaluation network, and optimize the error function to obtain the optimization result.

[0136] The estimation of the co-state of the evaluation network is expressed as:

[0137]

[0138] In the formula, For X(k),ω AFSEstimation of the costate equation for the (l+1)th iteration of active front wheel steering under the given state. For X(k),ω DYC Estimate the costate equation of the direct yaw moment in the (l+1)th iteration under the given condition. Let ω be the augmented state variable at time k. AFS ω represents the weight coefficients of the intermediate layer of the active front wheel steering neural network. DYC ω represents the weight coefficients of the intermediate layer in a direct yaw moment neural network. AFS(l+1) ω is the weighting coefficient for the (l+1)th iteration of the direct yaw moment. DYC(l+1) φ is the weighting coefficient for the (l+1)th iteration of the direct yaw moment. AFS(l+1) Let φ be the basis function for the (l+1)th iteration of the active front wheel steering. DYC(l+1) It is the basis function for the (l+1)th iteration of the direct yaw moment.

[0139] The expression for the target costate is as follows:

[0140]

[0141] In the formula, D AFS D is the weight matrix for active front wheel steering. DYC Q is the weight matrix for the direct yaw moment. AFS Q represents the tracking error of active front wheel steering. DYC For the tracking error of the direct yaw moment, λ AFS(l+1) (X(k)) is the co-state equation for the (l+1)th iteration of the active front wheel steering in state X(k), λ DYC(l+1) (X(k)) is the co-state equation for the (l+1)th iteration of the direct yaw moment in state X(k), γ AFS γ is a discount factor for the active front steering performance index. DYC This is a discount factor for the direct yaw moment performance index.

[0142] The expression for the error function of the evaluation network is:

[0143]

[0144] In the formula, e λAFS(l+1) (k) is the error function of the (l+1)th iteration of the active front wheel steering at time k, e λDYC(l+1) (k) is the error function of the (l+1)th iteration of the direct yaw moment at time k.

[0145] 3) Use a decision tree model to determine the driving conditions of the data collected in step 1) to obtain vehicle driving condition data;

[0146] The decision tree model is used for classification and regression tasks. It reads data from an ideal operational database; using the Gini index method, it selects a feature as the splitting criterion for each node, and splits the data in the ideal operational database into subsets based on the selected feature; it repeats the Gini index method for each subset, selecting the best feature for splitting, and continues splitting the data until a stopping condition is met; for new vehicle state data and road state data, it starts from the root node and proceeds according to the comparison results of feature values ​​until a leaf node is reached, where the value of the leaf node is the driving condition result.

[0147] 4) Using the data collected in step 1), the optimization results in step 2), and the vehicle driving condition data in step 3), game theory optimization is performed to obtain the actual output of active front wheel steering and the actual output of direct yaw torque; the specific steps are as follows:

[0148] 41) Using the data collected in step 1), the optimization results in step 2), and the vehicle driving condition data in step 3), calculate the state output equation of the closed-loop Nash equilibrium and the state equation of the reference signal system.

[0149] 42) Calculate the tracking error e of active front wheel steering. AFS (k) and the tracking error of the direct yaw moment e DYC (k);

[0150] 43) Calculate the cost function to enable the vehicle to track the reference trajectory while minimizing energy consumption;

[0151] 44) Calculate the state-space equation and performance index function of the augmented game system;

[0152] 45) Calculate the game output of the weighting factors.

[0153] The state output equation of the closed-loop Nash equilibrium in step 41) is:

[0154]

[0155] In the formula, x(k+1) is the state variable at time k+1, x(k) is the state variable at time k, and u AFS (k) represents the input for active front wheel steering at time k, u DYC (k) represents the input quantity of the direct yaw moment at time k, A is the coefficient matrix of the state variables, B AFS B is the coefficient matrix for the active front wheel steering input. DYC C is the coefficient matrix for the direct yaw moment input. AFS C is the output matrix for active front wheel steering. DYC The output matrix for the direct yaw moment, y AFS(k) represents the actual output of the active front wheel steering at time k, y DYC (k) represents the actual output of the direct yaw moment at time k;

[0156] The state equation of the reference signal system is:

[0157]

[0158] In the formula, The state of the reference signal system at time k+1, Let r be the state of the reference signal system at time k. AFS (k) represents the desired output of the active front wheel steering reference signal system at time k, r DYC (k) represents the desired output of the direct yaw moment reference signal system at time k, F is the system matrix corresponding to the reference signal system, and G... AFS G is the output matrix of the active front wheel steering reference signal system system matrix. DYC This is the output matrix of the system matrix for the direct yaw moment reference signal system.

[0159] In step 42), the tracking error of active front wheel steering and direct yaw moment is defined as follows:

[0160]

[0161] In the formula, e AFS (k) represents the tracking error of the active front wheel steering at time k, e DYC (k) represents the tracking error of the direct yaw moment at time k.

[0162] The cost function in step 43) is:

[0163]

[0164] in, for The cost function of active front wheel steering under certain conditions. for The cost function of the direct yaw moment under the condition, r AFS (e AFS (i),u AFS (i)) is e AFS (i),u AFS (i) The desired output of the active front wheel steering reference signal system under the given state, r DYC (e DYC (i),u DYC (i)) is e DYC (i),u DYC (i) The desired output of the direct yaw moment reference signal system at time k under state (i), rAFS (e AFS (i),u AFS (i)) and r DYC (e DYC (i),u DYC (i)) is represented as follows:

[0165]

[0166] In the formula, R AFS R is the control cost weight matrix for active front wheel steering. DYC This is the control cost weight matrix for the direct yaw moment.

[0167] The state-space equation and performance index function of the augmented game system in step 44) are as follows:

[0168]

[0169]

[0170] In the formula, Let be the augmented state variables at time k+1. Let k be the augmented state variable at time k. This is the coefficient matrix of the augmented state variables. This is the coefficient matrix for the augmented active front wheel steering input. J is the coefficient matrix of the augmented direct yaw moment input. AFS (X(k)) is the cost function of active front wheel steering in state X(k), J DYC (X(k)) is the cost function of the direct yaw moment in the state of X(k).

[0171] In step 45), the closed-loop Nash strategy is based on dynamic non-cooperative game theory. It holds if and only if the following recursive relation exists:

[0172]

[0173] in:

[0174]

[0175] In the formula, J at time k AFS (X(k)) intermediate variable, J at time k DYC (X(k)) intermediate variable, Let k be the optimal value for active front wheel steering. This represents the optimal value of the direct yaw moment at time k.

[0176] 5) Using the actual output of the active front wheel steering and the actual output of the direct yaw moment obtained in step 4), calculate the control commands for the active front wheel steering and the direct yaw moment; specifically including:

[0177] 51) Based on the actual output of the active front wheel steering and the actual output of the direct yaw moment in step 4), calculate the input estimate of the active front wheel steering in the lth iteration and the input estimate of the direct yaw moment in the lth iteration.

[0178] 52) Calculate the target value for the l-th iteration of the execution network;

[0179] 53) The error function of the execution network is obtained by taking the difference between the input estimate of the l-th iteration and the target value of the execution network;

[0180] 54) Adopting the adaptive rule of gradient descent, the weights of the execution network are updated, and the optimal control commands for active front wheel steering and direct yaw moment are generated. The vehicle is then controlled according to the control commands.

[0181] In step 51), based on the actual output of the active front wheel steering and the actual output of the direct yaw moment output in step 4), the input estimates for the l-th iteration of the active front wheel steering and the input estimates for the l-th iteration of the direct yaw moment are calculated.

[0182]

[0183] in, For X(k),θ AFS The input estimation for the l-th iteration of active front wheel steering under the given conditions. For X(k),θ DYC The input estimate of the direct yaw moment in the l-th iteration under the given state, θ AFS θ represents the weight coefficients of the intermediate layer of the active front wheel steering neural network. DYC ψ represents the weight coefficients of the intermediate layer in a direct yaw moment neural network. AFS(l) (X(k)) is the basis function for the l-th iteration of active front wheel steering in state X(k), ψ DYC(l) (X(k)) is the basis function of the l-th iteration of the direct yaw moment in the state of X(k).

[0184] Wherein, the target value for the l-th iteration of the network in step 52) is expressed by the following formula:

[0185]

[0186] Among them, u AFS(l)(X(k)) is the input quantity for the l-th iteration of active front wheel steering in state X(k); u DYC(l) (X(k)) is the input quantity of the direct yaw moment in the l-th iteration under the state of X(k); λ AFS(l) (X(k+1)) is the co-state equation for the l-th iteration of active front wheel steering in state X(k+1), λ DYC(l) (X(k+1)) is the co-state equation for the l-th iteration of the direct yaw moment in the state X(k+1).

[0187] In step 53), the error function of the execution network is obtained based on the target value of the execution network as follows:

[0188]

[0189] Among them, e uAFS ( l) (k) represents the target value of the active front wheel steering network in the l-th iteration, e uDYC ( l) (k) is the target value of the network for the l-th iteration of the direct yaw moment.

[0190] 6) Complete vehicle control according to the control instructions obtained in step 5).

[0191] This invention has many specific applications. The above description is only a preferred embodiment of this invention. It should be noted that for those skilled in the art, several improvements can be made without departing from the principle of this invention, and these improvements should also be considered within the scope of protection of this invention.

Claims

1. A cyber-physical system for an electro-hydraulic drive-by-wire chassis, characterized in that, include: Electro-hydraulic drive-by-wire chassis big data unit, vehicle-road-cloud traffic information unit, electro-hydraulic drive-by-wire chassis domain control unit, vehicle condition unit, road condition unit, and execution unit; The vehicle condition unit includes: a vehicle condition storage module and a vehicle condition sensor group module; The vehicle condition sensor module is used to collect vehicle status data and send the collected data to the vehicle status storage module. The vehicle status storage module is used to receive and store the data sent by the vehicle status sensor group module, and send the data to the first information storage module in the vehicle-road-cloud traffic information unit; The road condition unit includes: a road condition storage module and a road condition sensor group module; The road condition sensor module is used to collect road condition data and send the collected data to the road condition storage module. The road condition storage module is used to receive and store the data sent by the road condition sensor group module, and send the data to the first information storage module in the vehicle-road-cloud traffic information unit; The vehicle-road-cloud traffic information unit includes: a first information storage module and a game optimization module; The first information storage module transmits data and stores data with the electro-hydraulic drive-by-wire chassis big data unit, vehicle condition unit, road condition unit, and game optimization module, respectively. The game optimization module is used to optimize the active front wheel steering and direct yaw moment through game theory. The optimization results are the actual output of the active front wheel steering and the actual output of the direct yaw moment, and are sent to the electro-hydraulic drive-by-wire chassis big data unit. The electro-hydraulic drive-by-wire chassis big data unit includes: a data management module, a machine learning module, and an evaluation network module; The data management module transmits data and stores data with the first information storage module, the electro-hydraulic drive-by-wire chassis domain control unit, the evaluation network module, and the machine learning module. The evaluation network module is used to estimate the co-state of the evaluation network of the data collected by the vehicle condition sensor group module and the road condition sensor group module, calculate the target co-state of the data, and obtain the error function of the evaluation network by subtracting the estimated result of the co-state of the evaluation network from the target co-state. The error function of the evaluation network is optimized and the optimization result is sent to the data management module and the first information storage module. The machine learning module is used to determine the driving conditions and transmit the vehicle driving condition data to the data management module and the first information storage module. The electro-hydraulic drive-by-wire chassis domain control unit includes: a second information storage module and an execution network module; The second information storage module transmits data and stores data with the data management module, the execution unit, and the execution network module, respectively. The execution network module is used to generate control commands based on the actual output of active front wheel steering and the actual output of direct yaw torque data in the second information storage module, and send the control commands to the second information storage module. The execution unit includes: a drive-by-wire chassis steering module and a hub motor drive module; The drive-by-wire chassis steering module is used to perform active front wheel steering based on control commands generated by the execution network module; The hub motor drive module is used to execute direct yaw torque according to the control commands generated by the execution network module.

2. A game-theoretic optimization method for a cyber-physical system of an electro-hydraulic drive-by-wire chassis, based on the system described in claim 1, characterized in that, The steps are as follows: 1) Collect vehicle status data and road status data; 2) Estimate the co-state of the evaluation network for the data collected in step 1), calculate the target co-state of the data, calculate the difference between the estimated co-state of the evaluation network and the target co-state to obtain the error function of the evaluation network, and optimize the error function to obtain the optimization result. 3) Determine the driving conditions of the data collected in step 1) to obtain vehicle driving condition data; 4) Using the data collected in step 1), the optimization results in step 2), and the vehicle driving condition data in step 3), game optimization is performed to obtain the actual output of active front wheel steering and the actual output of direct yaw torque. 5) Calculate the control commands for active front wheel steering and direct yaw moment using the actual output of active front wheel steering and the actual output of direct yaw moment obtained in step 4). 6) Complete vehicle control according to the control instructions obtained in step 5).

3. The method according to claim 2, characterized in that, The estimation of the co-state of the evaluation network in step 2) is expressed as follows: In the formula, For X(k),ω AFS Estimation of the costate equation for the (l+1)th iteration of active front wheel steering under the given state. For X(k),ω DYC Estimate the costate equation of the direct yaw moment in the (l+1)th iteration under the given condition. Let ω be the augmented state variable at time k. AFS ω represents the weight coefficients of the intermediate layer of the active front wheel steering neural network. DYC ω represents the weight coefficients of the intermediate layer in a direct yaw moment neural network. AFS(l+1) ω is the weighting coefficient for the (l+1)th iteration of the direct yaw moment. DYC(l+1) φ is the weighting coefficient for the (l+1)th iteration of the direct yaw moment. AFS(l+1) Let φ be the basis function for the (l+1)th iteration of the active front wheel steering. DYC(l+1) It is the basis function for the (l+1)th iteration of the direct yaw moment.

4. The method according to claim 2, characterized in that, The expression for the target co-state in step 2) is as follows: Where D AFS D is the weight matrix for active front wheel steering. DYC Q is the weight matrix for the direct yaw moment. AFS Q represents the tracking error of active front wheel steering. DYC For the tracking error of the direct yaw moment, λ AFS(l+1) (X(k)) is the co-state equation for the (l+1)th iteration of the active front wheel steering in state X(k), λ DYC(l+1) (X(k)) is the co-state equation for the (l+1)th iteration of the direct yaw moment in state X(k), γ AFS γ is a discount factor for the active front steering performance index. DYC This is a discount factor for the direct yaw moment performance index.

5. The method according to claim 2, characterized in that, The specific steps of step 4) are as follows: 41) Using the data collected in step 1), the optimization results in step 2), and the vehicle driving condition data in step 3), calculate the state output equation of the closed-loop Nash equilibrium and the state equation of the reference signal system. 42) Calculate the tracking error e of active front wheel steering. AFS (k) and the tracking error e of the direct yaw moment DYC (k); 43) Calculate the cost function to enable the vehicle to track the reference trajectory while minimizing energy consumption; 44) Calculate the state-space equation and performance index function of the augmented game system; 45) Calculate the game output of the weighting factors.

6. The method according to claim 5, characterized in that, The state output equation of the closed-loop Nash equilibrium in step 41) is: In the formula, x(k+1) is the state variable at time k+1, x(k) is the state variable at time k, and u AFS (k) represents the input for active front wheel steering at time k, u DYC (k) represents the input quantity of the direct yaw moment at time k, A is the coefficient matrix of the state variables, B AFS B is the coefficient matrix for the active front wheel steering input. DYC C is the coefficient matrix for the direct yaw moment input. AFS C is the output matrix for active front wheel steering. DYC The output matrix for the direct yaw moment, y AFS (k) represents the actual output of the active front wheel steering at time k, y DYC (k) represents the actual output of the direct yaw moment at time k; The state equation of the reference signal system is: In the formula, The state of the reference signal system at time k+1, Let r be the state of the reference signal system at time k. AFS (k) represents the desired output of the active front wheel steering reference signal system at time k, r DYC (k) represents the desired output of the direct yaw moment reference signal system at time k, F is the system matrix corresponding to the reference signal system, and G... AFS G is the output matrix of the active front wheel steering reference signal system system matrix. DYC This is the output matrix of the system matrix for the direct yaw moment reference signal system.

7. The method according to claim 5, characterized in that, In step 42), the tracking error of active front wheel steering and direct yaw moment is defined as follows: In the formula, e AFS (k) represents the tracking error of the active front wheel steering at time k, e DYC (k) represents the tracking error of the direct yaw moment at time k.

8. The method according to claim 5, characterized in that, In step 45), based on dynamic non-cooperative game theory, the closed-loop Nash strategy... It holds if and only if the following recursive relation exists: in: In the formula, J at time k AFS (X(k)) intermediate variable, J at time k DYC (X(k)) intermediate variable, Let k be the optimal value for active front wheel steering. This represents the optimal value of the direct yaw moment at time k.

9. The method according to claim 2, characterized in that, Step 5) specifically includes: 51) Based on the actual output of the active front wheel steering and the actual output of the direct yaw moment in step 4), calculate the input estimate of the active front wheel steering in the lth iteration and the input estimate of the direct yaw moment in the lth iteration. 52) Calculate the target value for the l-th iteration of the execution network; 53) The error function of the execution network is obtained by taking the difference between the input estimate of the l-th iteration and the target value of the execution network; 54) Adopting the adaptive rule of gradient descent, the weights of the execution network are updated, and the optimal control commands for active front wheel steering and direct yaw moment are generated. The vehicle is then controlled according to the control commands.

10. The method according to claim 9, characterized in that, In step 51), based on the actual output of the active front wheel steering and the actual output of the direct yaw moment output in step 4), the input estimates for the l-th iteration of the active front wheel steering and the input estimates for the l-th iteration of the direct yaw moment are calculated: in, For X(k),θ AFS The input estimation for the l-th iteration of active front wheel steering under the given conditions. For X(k),θ DYC The input estimate of the direct yaw moment in the l-th iteration under the given state, θ AFS θ represents the weight coefficients of the intermediate layer of the active front wheel steering neural network. DYC ψ represents the weight coefficients of the intermediate layer in a direct yaw moment neural network. AFS(l) (X(k)) is the basis function for the l-th iteration of active front wheel steering in state X(k), ψ DYC(l) (X(k)) is the basis function of the l-th iteration of the direct yaw moment in the state of X(k).

Images