A full-vector power chassis vehicle drive and brake redundancy collaborative control method

Abstract

The present invention discloses a drive and brake redundancy collaborative control method for a full-vector power chassis vehicle, which relates to the technical field of automatic drive and brake control of vehicles. It solves the problems that the drive and brake control strategies of existing full-vector power chassis vehicles are not comprehensive enough and cannot well reduce the computing power requirements. The upper layer of the present invention calculates the longitudinal torque and yaw torque required by the vehicle, and the lower layer distributes the longitudinal torque and yaw torque sent by the upper layer controller by combining rules and look-up tables. The upper and lower layer actuators neural networks are trained by using intelligent evolutionary methods to achieve reinforcement learning control from vehicle state to command output, which simplifies the "signal acquisition - calculation and processing - output execution" and other links of the control system to the greatest extent, especially the hysteresis effect of the calculation and processing link, and improves the control efficiency and accuracy; to ensure the economy, safety and stability during the whole life cycle operation of the vehicle.

Description

Technical Field

[0001] The present invention relates to the technical field of automotive drive and brake control, and particularly relates to a full-vector power chassis vehicle drive and brake redundancy collaborative control method. Background Art

[0002] Currently, electric vehicles are the focus of research in the automotive industry, and full-vector power chassis vehicles are a type of electric vehicle that has received particular attention. They have the advantage of independent drive and brake for each wheel, and can better control the longitudinal force of each wheel's tire to optimize vehicle economy and control yaw stability. Moreover, when a failure occurs, precise control of each motor and brake can better reduce the harm and losses caused by the failure.

[0003] During vehicle driving, if drive or brake failure suddenly occurs and fault tolerance control is not carried out in a timely manner, it will seriously affect driving safety. However, the existing economy optimization or stability control strategies for full-vector power chassis vehicles are generally designed only for vehicles in normal driving states, and there is relatively little research on fault tolerance control strategies under the above-mentioned failure conditions. Most of the existing fault tolerance control strategies under failure conditions only target single drive failure or brake failure, without fully integrating the overall drive and brake coordination, which has limitations and affects vehicle operation stability and driving safety.

[0004] For example, some fault tolerance control strategies under failure conditions are as follows:

[0005] A drive system fault tolerance control strategy for a distributed drive electric vehicle disclosed in Patent No. 201910126899.4 calculates the expected yaw moment through a sliding mode variable structure control strategy in the upper layer control; in the lower layer control, the distribution of wheel torques is calculated by combining the sliding mode variable structure control with the expected yaw moment, and torque re-distribution is performed on the normally operating wheels to ensure vehicle driving stability. However, it still has the following deficiencies: both the upper layer calculation and the lower layer distribution use the sliding mode algorithm, which requires high computing power; the expression method for torque reconstruction after failure is too simple, only giving the control method after single-wheel failure, without considering multiple failure situations; it only expresses the control method for drive part failure and does not have an expression for the brake system.

[0006] A control method for a distributed drive electric vehicle drive system based on failure states disclosed in Patent No. 202010662004.1 classifies and studies the failure situations of the distributed drive electric vehicle drive system, and further refines the failure control according to the corresponding failure torque re-distribution control strategy. However, it still has the following deficiencies: it does not mention the strategy expression in the non-failure state, only focuses on the coping strategies in the drive failure state, and only expresses the control method for drive part failure and does not have an expression for the brake system.

[0007] A hub motor-driven vehicle electro-mechanical-hydraulic redundant braking system and control method disclosed in Patent No. 201711285373.8 preferentially uses the hub motor for regenerative braking. The electro-mechanical braking system is used for driving braking compensation and parking braking, and the hydraulic braking system is used for failure compensation. The three braking systems cooperate to judge the braking mode based on the comprehensive road conditions and the state of the power battery, solve the total braking torque demand, reasonably distribute the braking torques of the four wheels, and the hub motor controller, the electro-mechanical braking system controller, and the electromechanical brake booster cooperate to perform braking control to jointly achieve the braking target and ensure driving stability. However, it still has the following deficiencies: only the control method for partial braking failure is expressed, without the expression of the drive system; the control strategy under brake failure is not considered; the form of braking torque distribution calculated in real time after real-time monitoring of the state has high requirements for computing power.

[0008] It can be seen that the existing control strategies for full-vector power chassis vehicles cannot effectively control the vehicle when abnormal driving and braking occur, and do not consider the problem that excessive computing power in the control process may lead to control time delay. Summary of the Invention

[0009] In order to solve the problems existing in the above-mentioned prior art, the present invention provides a drive-brake redundant cooperative control method for a full-vector power chassis vehicle, which solves the problems that the drive-brake control strategy of the existing full-vector power chassis vehicle is not comprehensive enough and cannot determine the control strategy with low computing power.

[0010] A drive-brake redundant cooperative control method for a full-vector power chassis vehicle includes a hierarchical control method and an intelligent evolution method.

[0011] The hierarchical control method includes calculating the expected torque according to the vehicle driving information, and then selecting the corresponding torque distribution strategy to distribute the expected torque by combining rules and look-up tables according to the vehicle state. The vehicle state includes the driving state and the braking state. The driving state includes normal driving and driving failure, and the braking state includes normal braking and braking failure. The expected torque includes the expected longitudinal torque and the expected yaw torque.

[0012] The intelligent evolution method includes making a data set based on the relevant information in the control process of the hierarchical control method, including the vehicle driving information and the corresponding determined torque distribution strategy, to train the neural network model to obtain a strategy generator. The strategy generator can directly output the expected torque and the torque distribution strategy according to the vehicle driving information to control the vehicle.

[0013] The hierarchical control method and the intelligent evolution method are used alternately.

[0014] Moreover, considering both the driving state and the braking state, the range that can be addressed is more comprehensive than that of single driving control and braking control, and the generated strategy is also more perfect.

[0015] Furthermore, the alternating use means switching to the intelligent evolution method at the first switching opportunity and switching to the hierarchical control method at the second switching time. The first switching opportunity means that the accuracy of the expected torque and torque distribution strategy output by the strategy generator reaches the expectation, and the second switching opportunity means that the vehicle driving mileage or duration reaches the threshold after adopting the intelligent evolution method.

[0016] First, use the hierarchical control method to control the vehicle, realize data acquisition and training while ensuring the control accuracy. After obtaining enough data, train and generate a qualified strategy generator, and then the strategy generator can directly replace the complex strategy determination process to quickly and timely generate corresponding strategies, reducing the computing power consumption; when the vehicle driving mileage reaches a certain value, at this time the vehicle state itself changes, and the accuracy of the strategy generated by the original strategy generator will decrease. At this time, re-enable the hierarchical control method to control the vehicle, collect the data set again to train the strategy generator until the accuracy of the strategy generated by the strategy generator reaches the expected level, and then re-enable the strategy generator, and so on, so as to minimize the computing power requirements for generating fault-tolerant strategies while ensuring the driving stability and safety of the vehicle.

[0017] Furthermore, when the braking is normal, the expected torque is distributed to the brake actuator by using the braking torque distribution MAP. The brake actuator includes an electric brake actuator and a hydraulic brake actuator. The distribution rule of the braking torque distribution MAP is: if the electric brake actuator meets the expected torque, pure electric braking is adopted; if the electric brake actuator cannot meet the expected torque, the hydraulic brake actuator is used to make up for the insufficient part of the electric brake actuator.

[0018] Furthermore, when the braking fails, the expected torque is distributed to the brake actuator by using the braking torque distribution rule. The brake actuator includes an electric brake actuator and a hydraulic brake actuator. The braking torque distribution rule is: if the front axle or the rear axle of the electric brake actuator fails, the electric brake corresponding to the axle without failure and the hydraulic brake actuator are used for braking; if both the front axle and the rear axle of the electric brake actuator fail, the braking is completely carried out by the hydraulic brake actuator; if the front axle or the rear axle of the hydraulic brake actuator fails, the hydraulic brake corresponding to the axle without failure and the electric brake actuator are used for braking; if both the front axle and the rear axle of the hydraulic brake actuator fail, the braking is completely carried out by the electric brake actuator. If the electric brake actuator is not sufficient to provide the expected torque, the vehicle speed is reduced to reduce the braking demand torque and stop the vehicle.

[0019] Further, when the drive is normal, the desired torque is distributed using the drive torque distribution MAP. The process of forming the drive torque distribution MAP is as follows:

[0020] Define the torque distribution coefficient λ:

[0021]

[0022] In the formula, T d is the desired torque, T fl , T fr are the torques distributed to the left front wheel and the right front wheel, T rl , T rr are the torques distributed to the left rear wheel and the right rear wheel, and λ is the front and rear axle torque distribution coefficient;

[0023] When λ = 0, it indicates rear-wheel drive. When λ = 1, it indicates front-wheel drive. The total drive efficiency is obtained as shown in the following formula:

[0024]

[0025] In the formula, η fl (T fl , n * ) is the front-wheel efficiency of the left front-wheel torque T fl at the rotational speed n * , η fr (T fr , n * ) is the front-wheel efficiency of the torque T fr at the rotational speed n * , η rl (T rl , n * ) is the rear-wheel efficiency of the torque T rl at the rotational speed n * , η rr (T rr , n * ) is the rear-wheel efficiency of the torque T rr at the rotational speed n * , η is the total drive efficiency of the vehicle, T d is the driver's desired torque, and η fl , η fr , η rl , η rr are the motor speeds of the left front, right front, left rear, and right rear wheels respectively;

[0026] Perform a traversal loop calculation according to the above formula to obtain the torque distribution coefficient λ with the optimal efficiency between different total desired torques and motor speeds, that is, obtain the drive torque distribution MAP with the desired torque T d and the motor speed as the input and the torque distribution coefficient λ as the output.

[0027] Further, when the drive fails, first determine the failure location and then process it according to the established rules. The failure locations of the drive failure include single motor failure, coaxial dual motor failure, same-side dual motor failure, different-axis and different-side dual motor failure, and triple motor failure. If the failure location is single motor failure or coaxial dual motor failure, the motor corresponding to the non-failed axis provides the desired torque. If the failure location is different-axis and different-side dual motor failure, the non-failed motors provide the desired torque. If the failure location is same-side dual motor failure or triple motor failure, the non-failed motors provide the desired torque, and at the same time, keep the output torque of a single non-failed motor less than the calibrated value to ensure the vehicle travels at a low speed and avoid the influence of unbalanced left and right torques on safety.

[0028] Further, the strategy generator includes a torque generation module and a torque distribution module. The torque generation is used to directly output the desired torque, and the torque distribution module is used to directly determine the torque distribution strategy according to the desired torque and the vehicle state;

[0029] For the torque generation module, its state space and execution actions can be defined as:

[0030]

[0031] where α is the accelerator pedal signal, β is the brake pedal signal, F x is the desired longitudinal torque, and ΔM is the desired yaw torque; is the longitudinal vehicle speed; V y is the lateral vehicle speed; ω z is the yaw angular velocity;

[0032] For the torque distribution module, its state space and execution actions can be defined as:

[0033]

[0034] where E Tij(i＝1，2；j＝1，2) is the state of the four motors, E Bij(i＝1，2；j＝1，2) is the state of the four hydraulic brakes, T Tij(i＝1，2；j＝1，2) is the torque command output to the four motors, and T Bij(i＝1，2；j＝1，2) is the braking torque command output to the four hydraulic brakes.

[0035] Further, in the calculation process of the desired longitudinal torque, the desired speed is used as the control target. In the calculation process of the desired yaw torque, the yaw angular velocity and the center of mass deflection angle are used as the control targets. The desired yaw torque is determined by the vehicle attitude. When the vehicle is not unstable, the desired yaw torque is 0, and at this time, the desired torque is the desired longitudinal torque; when the vehicle has a tendency to become unstable, the desired torque includes both the desired driving torque and the desired yaw torque at this time.

[0036] Further, the vehicle driving information includes the vehicle operation state feedback by the vehicle pose sensor and the driver operation signal, that is, the high-order intelligent driving instruction includes the accelerator pedal signal and the brake pedal signal.

[0037] Further, the hierarchical control method includes upper-layer control and lower-layer control. The upper-layer control includes calculating the desired torque according to the vehicle driving information; the lower-layer control includes judging the state of the drive actuator or the brake actuator, and then selecting the corresponding torque distribution strategy to distribute the desired torque by combining rules and look-up tables according to the state.

[0038] The beneficial effects of the present invention include:

[0039] (1) Considering the coordination of the drive system actuator and the brake system actuator of the full-vector power chassis vehicle, it is more perfect and comprehensive than the existing rule-based drive and brake fault handling methods or separate drive stability control strategies.

[0040] (2) Under the premise of integrating drive and brake failure fault tolerance control, the vehicle drive economy, vehicle stability, brake energy recovery utilization efficiency, and brake smoothness can be comprehensively considered.

[0041] (3) Using a control strategy that combines rules and MAP look-up tables can save computing power while ensuring the economy, safety, and stability during the entire life cycle operation of the vehicle.

[0042] (4) Training the upper and lower layer actuator neural networks using intelligent evolutionary methods realizes reinforcement learning control from vehicle state to command output, greatly simplifies the "signal acquisition - calculation processing - output execution" and other links of the control system, especially the hysteresis effect in the calculation processing link, and improves the control efficiency and accuracy; Description of the Drawings

[0043] Figure 1 It is a flowchart of a drive and brake redundancy collaborative control method for a full-vector power chassis vehicle. Specific Embodiments

[0044] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. Therefore, the detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application to be protected, but only represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative efforts shall fall within the scope of protection of the present application.

[0045] A full-vector dynamic chassis vehicle drive and brake redundancy collaborative control method, including a hierarchical control method and an intelligent evolution method,

[0046] The hierarchical control method includes calculating the desired torque according to the vehicle driving information, and then selecting the corresponding torque distribution strategy according to the vehicle state to distribute the desired torque. The vehicle state includes the driving state and the braking state. The driving state includes normal driving and driving failure. The braking state includes normal braking and braking failure. The desired torque includes the desired longitudinal torque and the desired yaw torque;

[0047] The intelligent evolution method includes making a data set based on the relevant information in the control process of the hierarchical control method, including the vehicle driving information and the corresponding determined torque distribution strategy, to train a neural network model to obtain a strategy generator. The strategy generator can directly output the torque distribution strategy according to the vehicle driving information to control the vehicle; The hierarchical control method and the intelligent evolution method are used alternately.

[0048] Considering both the driving state and the braking state, compared with single driving control and braking control, the range that can be dealt with is more comprehensive, and the generated strategy is also more perfect.

[0049] Specifically, at a specific moment, the vehicle can only be in a specific state in the driving state or the braking state. When the vehicle is in the driving state, the desired torque is the desired driving torque and the desired yaw torque. When the vehicle is in the braking state, the desired torque is the desired braking torque and the desired yaw torque.

[0050] When in the driving state, the desired torque is calculated by the driving motion tracking controller and distributed by the driving distribution controller. When in the braking state, the desired torque is calculated by the braking motion tracking controller and distributed by the braking distribution controller.

[0051] When in the driving state, the driving motion tracking controller will select a strategy according to the driving motor failure state. When there is no motor failure, the torque distribution between the front and rear axles adopts the form of MAP look-up table to improve the vehicle driving economy. When a motor failure occurs, torque distribution and torque reconstruction are carried out according to the established rules to ensure vehicle stability and power performance.

[0052] When in the braking state, the braking motion tracking controller selects strategies according to the failure states of the braking actuators (electric motor braking and hydraulic braking). When there is no failure of the braking actuators, the electro-hydraulic braking distribution ratio adopts the form of MAP look-up table to improve the vehicle energy recovery efficiency. When a braking actuator fails, the braking torque is distributed and reconstructed according to the established rules to ensure vehicle stability and braking performance. After the relevant MAP is first written into the controller, DQN learning training and periodic iteration are continuously carried out during operation.

[0053] In another embodiment, the hierarchical control method and the intelligent evolution method are used alternately; the alternate use means switching to the intelligent evolution method at the first switching opportunity and switching to the hierarchical control method at the second switching time. The first switching opportunity refers to that the accuracy rates of the expected torque and the torque distribution strategy output by the strategy generator reach the expectation, and the second switching opportunity refers to that the vehicle driving mileage or duration reaches the threshold after adopting the intelligent evolution method.

[0054] First, use the hierarchical control method to control the vehicle, realize data acquisition and training while ensuring the control accuracy. After obtaining enough data, train and generate a qualified strategy generator, and then directly use the strategy generator to replace the complex strategy determination process, quickly and timely generate corresponding strategies, and reduce the computing power consumption. When the vehicle driving mileage reaches a certain value, at this time the vehicle state itself changes, and the accuracy rate of the strategy generated by the original strategy generator will decrease. At this time, re-enable the hierarchical control method to control the vehicle, collect the data set again to train the strategy generator until the accuracy rate of the strategy generated by the strategy generator reaches the expected level, and then re-enable the strategy generator, and so on, so as to minimize the computing power requirements for generating fault-tolerant strategies while ensuring vehicle driving stability and safety.

[0055] Specifically, since the state changes of the vehicle during the whole life cycle operation will affect the control accuracy of the reinforcement learning network, based on the vehicle driving mileage interval (not less than 10,000 kilometers) and the update time interval (not less than half a year) as the judgment conditions, after reaching the judgment conditions, re-adopt the hierarchical control method to take over the vehicle control in the "signal acquisition - calculation processing - output execution" link. After continuous control for a period of time to have sufficient data, adopt the intelligent evolution method to re-train the strategy generator, and after training, replace it in the same way to achieve evolutionary iteration.

[0056] In another embodiment, when the braking is normal, the expected torque is distributed to the braking actuators by using the braking torque distribution MAP. The braking actuators include an electric braking actuator and a hydraulic braking actuator. The distribution rule of the braking torque distribution MAP is: if the electric braking actuator meets the expected torque, pure electric braking is adopted; if the electric braking actuator cannot meet the expected torque, the hydraulic braking actuator is used to make up for the insufficient part of the electric braking actuator.

[0057] Specifically, in the braking execution part, in a fault-free state, the braking torque distribution among the braking actuators is determined by the braking torque distribution MAP (referred to as MAP_B for short). MAP_B is a distribution MAP based on the idea of economy. At this time, to maximize the energy recovery and energy-saving effect, if the electric braking torque meets the braking torque requirement, pure electric braking is performed; if the motor can participate in braking but the electric braking torque cannot meet the vehicle's total braking torque requirement, the hydraulic braking system is activated to make up for the insufficient total vehicle braking torque requirement with the hydraulic braking torque. Based on this idea, the total braking demand torque distribution coefficient is a parameter output by MAP_B. Define the total electric braking torque of the motor as T BM , the total hydraulic braking torque as T BH and the desired longitudinal torque as T b to satisfy the following relationship

[0058]

[0059] On this basis, to ensure safety and avoid the electric braking command exceeding the threshold, the electric braking torque output by each motor is T BM / 4 without adjustment. Each hydraulic brake provides an additional yaw moment ΔM by adjusting the left and right braking torques on the basis of outputting a braking torque of T BH / 4.

[0060] In another embodiment, when braking fails, the braking torque distribution rule is used to distribute the desired torque to the braking actuators. The braking actuators include electric braking actuators and hydraulic braking actuators. The braking torque distribution rule is as follows: if an electric brake fails in the front axle or the rear axle of the electric braking actuators, the electric brakes corresponding to the axle without failure and the hydraulic braking actuators are used for braking; if electric brakes fail in both the front axle and the rear axle, braking is completely performed by the hydraulic braking actuators; if a hydraulic brake fails in the front axle or the rear axle of the hydraulic braking actuators, the hydraulic brakes corresponding to the axle without failure and the electric braking actuators are used for braking; if hydraulic brakes fail in both the front axle and the rear axle, braking is completely performed by the electric braking actuators. If the electric braking actuators are not sufficient to provide the desired torque, the vehicle speed is reduced to reduce the braking demand torque and stop the vehicle.

[0061] Specifically, the electric braking failure mainly refers to three situations: the motor cannot output torque commands, the traction motor speed is too low (the braking energy recovery has restrictions on the motor speed. According to the power generation efficiency characteristics of the traction motor, when the traction motor participates in braking, if the speed is low, neither the motor power generation efficiency nor the motor control accuracy is conducive to electric braking. At the same time, the large torque at low speed will cause the motor to generate more heat; if the motor speed is too low, the motor efficiency is too low and the large heat generation will affect the motor life. Therefore, when the motor speed is lower than a certain value, it is regarded as electric braking failure), and the traction motor temperature is too high (the large heat generation will affect the motor life. Therefore, when the motor temperature is too high, it is regarded as electric braking failure).

[0062] When there is a fault state, for the electric braking fault, the four motors are divided into two groups: the two motors on the front axle are incorporated into the front axle motor group, and the two motors on the rear axle are incorporated into the rear axle motor group. To ensure stability during braking, it is not allowed that one motor in the same group brakes while the other does not brake. Only uniform braking or non-braking is allowed.

[0063] Taking the front axle as an example, if a certain motor on the front axle cannot participate in braking due to its speed, temperature, or a fault, then both motors on the front axle do not brake. At this moment, to ensure safety, the rear axle motors and the hydraulic brakes operate normally for braking, and the braking torque that the two front axle motors cannot provide is compensated by the front axle hydraulic brakes. The same treatment method is applied to the rear axle motor faults. If motors on both the front and rear axles cannot participate in braking, then all the braking torque and the additional yaw moment are provided by the hydraulic brakes.

[0064] Regarding the hydraulic braking failure, it mainly means that the hydraulic brake cannot complete the braking action.

[0065] Similarly, the two hydraulic brakes on the front axle are incorporated into the front axle hydraulic braking group, and the two hydraulic brakes on the rear axle are incorporated into the rear axle hydraulic braking group. To ensure stability during braking, it is not allowed that one hydraulic brake in the same group brakes while the other does not brake. Only uniform braking or non-braking is allowed.

[0066] Taking the front axle as an example, if a certain hydraulic brake on the front axle cannot participate in braking, then both hydraulic brakes on the front axle do not brake. At this moment, to ensure safety, the front axle motors, the rear axle motors, and the rear axle hydraulic brakes operate normally for braking, and the braking torque that the two front axle hydraulic brakes cannot provide is compensated by the rear axle hydraulic brakes. The treatment method for the rear axle hydraulic brake faults is similar. If hydraulic brakes on both the front and rear axles fail, since the electric braking cannot provide a large braking torque, in this state, to ensure safety, the driver should be reminded immediately and the vehicle should be decelerated and stopped in a timely manner.

[0067] In another embodiment, when the driving is normal, the desired torque is distributed using the drive torque distribution MAP (abbreviated as MAP_D). The formation process of the MAP_D is as follows:

[0068] Define the torque distribution coefficient λ:

[0069]

[0070] where T d is the desired longitudinal torque, T fl , T fr are the torques distributed to the left front wheel and the right front wheel, T rl , T rr are the torques distributed to the left rear wheel and the right rear wheel, and λ is the torque distribution coefficient between the front and rear axles;

[0071] When λ = 0, it indicates rear-wheel drive; when λ = 1, it indicates front-wheel drive. The total drive efficiency is shown by the following formula:

[0072]

[0073] where η fl (T fl , n * ) is the front-wheel efficiency at the left front-wheel torque T fl and rotational speed n * , η fr (T fr , n * ) is the front-wheel efficiency at the torque T fr and rotational speed n * , η rl (T rl , n * ) is the rear-wheel efficiency at the torque T rl and rotational speed n * , η rr (T rr , n * ) is the rear-wheel efficiency at the torque T rr and rotational speed n * , η is the total drive efficiency of the four drive motors, T d is the driver's desired torque, and η fl , η fr , η rl , η rr are the rotational speeds of the motors of the left front, right front, left rear, and right rear wheels respectively;

[0074] According to the above formula, perform a traversal loop calculation to obtain the torque distribution coefficient λ with the optimal efficiency between different total desired torques and motor speeds, that is, obtain MAP_D with the desired torque T d and motor speed as inputs and the torque distribution coefficient λ as the output.

[0075] Specifically, when the drive is normal, i.e., without failure, the drive torques corresponding to each wheel satisfy the following conditions:

[0076]

[0077] Where B is the wheelbase of the front and rear axles (generally considered equal), and r is the tire radius.

[0078] In another embodiment, when the drive fails, the failure location is first determined and then processed according to established rules. The failure locations of the drive failure include single-motor failure, coaxial dual-motor failure, same-side dual-motor failure, different-axis and different-side dual-motor failure, and triple-motor failure. If the failure location is single-motor failure or coaxial dual-motor failure, the motor corresponding to the non-failed axle provides the desired torque. If the failure location is different-axis and different-side dual-motor failure, the non-failed motor provides the desired torque. If the failure location is same-side dual-motor failure or triple-motor failure, the non-failed motor provides the desired torque, and at the same time, the output torque of a single non-failed motor is kept less than the calibration value to ensure low-speed vehicle driving and avoid the influence of left-right torque imbalance on safety.

[0079] Specifically, the following are examples of the processing methods corresponding to each failure location:

[0080]

[0081]

[0082] In another embodiment, the strategy generator includes a torque generation module and a torque distribution module. The torque generation is used to directly output the desired torque, and the torque distribution module is used to directly determine the torque distribution strategy according to the desired torque and the vehicle state;

[0083] For the torque generation module, its state space and execution actions can be defined as:

[0084]

[0085] In the formula, α is the driver's accelerator pedal signal (acceleration instruction of the intelligent driving system), β is the driver's brake pedal signal (braking instruction of the intelligent driving system), F x is the desired longitudinal torque, and ΔM is the desired yaw torque; is the longitudinal vehicle speed; V y is the lateral vehicle speed; ω z is the yaw angular velocity

[0086] For the torque distribution module, its state space and execution actions can be defined as:

[0087]

[0088] In the formula, ETij(i＝1，2；j＝1，2) is the state of four motors, E Bij(i＝1，2；j＝1，2) is the state of four hydraulic brakes, T Tij(i＝1，2；j＝1，2) is the torque command output to four motors, T Bij(i＝1，2；j＝1，2) is the braking torque command output to four hydraulic brakes. Specifically, when in the driving state, the four motors drive the corresponding wheels respectively, and when in the braking state, the four motors act as electric brakes to brake the corresponding wheels respectively.

[0089] In this embodiment, the reinforcement learning method adopts the DQN reinforcement learning method, and the specific content is as follows:

[0090] The agent of reinforcement learning interacts with the environment by using a way of selecting actions with the maximum future feedback value, and calculates the action Q value by defining the optimal action selection function Q(s, a);

[0091] Q(s, a) = max π E[R t / s t = s, a t = a / π]

[0092] where s is the state, a is the action executed in this state; π is the mapping of actions and states; s t is the state at time step t, a t is the action executed in state s t ; R t is the feedback value obtained by executing action a in state s, and its formula is as follows;

[0093]

[0094] where t is the time step; T is the total number of time steps when reaching the termination state; r t′ is the feedback value at the t'-th time step; γ represents the attenuation rate of the feedback value at each time step;

[0095] Deep learning uses the samples provided by reinforcement learning to learn all the action Q values in this state, and calculates the loss function L i (θ i ) as shown in Equation (5); further, by using the stochastic gradient descent method on the loss function to update the weights θ of the neural network approximator, the purpose of optimizing the Q neural network is achieved;

[0096] L i (θ i ) = E s，a～ρ(·) [(y i -Q(s, a; θ i )) 2 )

[0097] where Q(s, a; θ i ) is an estimate of Q(s, a); i is the number of iterations; s is the current state; a is the currently selected action; ρ(s, a) is the probability distribution between state s and action a; y i is Q(s, a) obtained according to the Bellman equation and can be expressed as in Equation (6);

[0098] y i = E s′ [r + γ max a′ Q i (s′, a′; θ i-1 ) / s, a]

[0099] where s′ is the next state; a′ is the next action.

[0100] In another embodiment, in the process of calculating the expected longitudinal torque, the expected speed is used as the control target, and in the process of calculating the expected yaw moment, the yaw angular velocity and the center-of-mass deflection angle are used as the control targets. The expected yaw moment is determined by the vehicle attitude. When the vehicle is not unstable, the expected yaw moment is 0, and at this time the expected torque is the expected longitudinal torque; when the vehicle has a tendency to be unstable, the expected torque includes both the expected driving torque and the expected yaw moment at this time.

[0101] In another embodiment, the vehicle driving information includes the vehicle running state feedback by the vehicle pose sensor and the driver operation signal, that is, the high-order intelligent driving instruction includes the driver's accelerator pedal signal (intelligent driving system acceleration instruction) and the driver's brake pedal signal (intelligent driving system braking instruction).

[0102] In another embodiment, the hierarchical control method includes upper-layer control and lower-layer control. The upper-layer control includes calculating the expected torque according to the vehicle driving information; the lower-layer control includes judging the state of the driving actuator or the braking actuator, and then selecting the corresponding torque distribution strategy according to the state to distribute the expected torque.

[0103] The above-described embodiments only represent the specific implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation to the protection scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the technical solution of the present application, several deformations and improvements can still be made, and these all belong to the protection scope of the present application.

Claims

1. A full-vector power chassis vehicle drive and brake redundancy collaborative control method, characterized in that It includes a hierarchical control method and an intelligent evolution method. The hierarchical control method includes calculating the expected torque according to the vehicle driving information, and then selecting a corresponding torque distribution strategy in a combined way of rules and look-up tables according to the vehicle state to distribute the expected torque. The vehicle state includes the driving state and the braking state. The driving state includes normal driving and driving failure. The braking state includes normal braking and braking failure. The expected torque includes the expected longitudinal torque and the expected yaw torque. The intelligent evolution method includes training a neural network model with a data set made according to the relevant information in the control process of the hierarchical control method, including the vehicle driving information and the corresponding determined torque distribution strategy, to obtain a strategy generator. The strategy generator can directly output the expected torque and the torque distribution strategy according to the vehicle driving information to control the vehicle. The hierarchical control method and the intelligent evolution method are used alternately. The alternate use means switching to the intelligent evolution method at the first switching opportunity and switching to the hierarchical control method at the second switching opportunity. The first switching opportunity means that the accuracy rate of the expected torque and the torque distribution strategy output by the strategy generator reaches the expectation. The second switching opportunity means that the vehicle driving mileage or duration reaches the threshold after using the intelligent evolution method. The strategy generator includes a torque generation module and a torque distribution module. The torque generation is used to directly output the expected torque. The torque distribution module is used to directly determine the torque distribution strategy according to the expected torque and the vehicle state. For the torque generation module, its state space and execution actions can be defined as: where α is the accelerator pedal signal, β is the brake pedal signal, F x is the desired longitudinal torque, and ΔM is the desired yaw torque; is the longitudinal vehicle speed; is the lateral vehicle speed; ω z is the yaw rate; For the torque distribution module, its state space and execution actions can be defined as: where E Tij(i＝1,2；j＝1,2) represents the states of four motors, and E Bij(i＝1,2；j＝1,2) represents the states of four hydraulic brakes. T Tij(i＝1,2；j＝1,2) is the torque command output to the four motors, and T Bij(i＝1,2；j＝1,2) is the braking torque command output to the four hydraulic brakes.

2. The full-vector power chassis vehicle drive and brake redundancy collaborative control method according to claim 1, wherein, When the braking is normal, the braking torque distribution MAP is used to distribute the expected torque. The braking actuator includes an electric braking actuator and a hydraulic braking actuator. The distribution rule of the braking torque distribution MAP is: if the electric braking actuator meets the expected torque, pure electric braking is used. If the electric braking actuator cannot meet the expected torque, the hydraulic braking actuator is used to make up for the insufficient part of the electric braking actuator.

3. A full-vector dynamic chassis vehicle drive and brake redundancy collaborative control method according to claim 1, characterized in that, When the braking fails, the braking torque distribution rule is used to distribute the expected torque to the braking actuator. The braking actuator includes an electric braking actuator and a hydraulic braking actuator. The braking torque distribution rule is: if the front axle or the rear axle of the electric braking actuator fails, the electric brake corresponding to the axle without failure and the hydraulic braking actuator are used for braking. If both the front axle and the rear axle of the electric braking actuator fail, the braking is completely carried out by the hydraulic braking actuator. If the front axle or the rear axle of the hydraulic braking actuator fails, the hydraulic brake corresponding to the axle without failure and the electric braking actuator are used for braking. If both the front axle and the rear axle of the hydraulic braking actuator fail, the braking is completely carried out by the electric braking actuator. If the electric braking actuator is not enough to provide the expected torque, the vehicle speed is reduced to reduce the braking demand torque and stop the vehicle.

4. A full-vector power chassis vehicle drive and brake redundancy collaborative control method according to claim 1, characterized in that, When the driving is normal, the driving torque distribution MAP is used to distribute the expected torque to the braking actuator. The formation process of the driving torque distribution MAP is as follows: Define the torque distribution coefficient λ: Wherein, T d is the desired torque, T fl , T fr are the torques distributed to the left front wheel and the right front wheel, T rl , T rr are the torques distributed to the left rear wheel and the right rear wheel, and λ is the front and rear axle torque distribution coefficient; When λ = 0, it represents rear-wheel drive, and when λ = 1, it represents front-wheel drive. The total drive efficiency is shown as follows: where η fl (T fl , n * ) is the front left wheel torque T fl at the rotational speed n * , and η fr (T fr , n * ) is the front right wheel torque T fr at the rotational speed n * , and η rl (T rl , n * ) is the rear left wheel torque T rl at the rotational speed n * , and η rr (T rr , n * ) is the rear right wheel torque T rr at the rotational speed n * , η is the total driving efficiency of the vehicle, T d is the driver's desired torque, and η fl , η fr , η rl , η rr are the rotational speeds of the front left, front right, rear left, and rear right wheel motors respectively; Traverse and loop calculate according to the above formula to obtain the torque distribution coefficient λ with the optimal efficiency between different total expected torques and motor speeds, that is, obtain the drive torque distribution MAP with the expected torque T d and motor speed as inputs and the torque distribution coefficient λ as the output.

5. A full-vector power chassis vehicle drive and brake redundancy collaborative control method according to claim 1, characterized in that, When drive failure occurs, first determine the failure location and then process it according to the established rules. The failure locations of the drive failure include single-motor failure, coaxial dual-motor failure, same-side dual-motor failure, different-axis and different-side dual-motor failure, and triple-motor failure. If the failure location is single-motor failure or coaxial dual-motor failure, the motor corresponding to the non-failed axis provides the desired torque. If the failure location is different-axis and different-side dual-motor failure, the non-failed motor provides the desired torque. If the failure location is same-side dual-motor failure or triple-motor failure, the non-failed motor provides the desired torque, and at the same time, the output torque of a single non-failed motor is kept less than the calibrated value.

6. A full-vector power chassis vehicle drive and brake redundancy collaborative control method according to any one of claims 1-5, characterized in that, In the calculation process of the desired longitudinal torque, the desired speed is used as the control target. In the calculation process of the desired yaw torque, the yaw angular velocity and the center-of-mass deflection angle are used as the control targets. The desired yaw torque is determined by the vehicle attitude. When the vehicle is not unstable, the desired yaw torque is 0, and at this time, the desired torque is the desired longitudinal torque; When the vehicle has a tendency to become unstable, the desired torque at this time includes both the desired drive torque and the desired yaw torque.

7. A full-vector dynamic chassis vehicle drive and brake redundancy collaborative control method according to any one of claims 1-5, characterized in that, The vehicle driving information includes the vehicle operation state feedback by the vehicle pose sensor and the driver operation signal, that is, the high-order intelligent driving instruction includes the accelerator pedal signal and the brake pedal signal.

8. A full-vector dynamic chassis vehicle drive and brake redundancy collaborative control method according to any one of claims 1-5, characterized in that The hierarchical control method includes upper-layer control and lower-layer control. The upper-layer control includes calculating the desired torque according to the vehicle driving information; the lower-layer control includes judging the state of the drive actuator or the brake actuator, and then selecting the corresponding torque distribution strategy to distribute the desired torque in a way that combines rules and look-up tables according to the state.

Images