Vehicle vertical-roll cooperative control method based on roll phase plane switching

Abstract

The invention discloses a vehicle vertical-roll cooperative control method based on roll phase plane switching, and the method comprises the steps: 1, building a six-degree-of-freedom steering-roll vehicle dynamics model which is used for describing the vertical vibration of sprung and unsprung masses and the roll, yaw and lateral movement of a vehicle; 2, designing a state space, an action space, a multi-mode reward function and a network structure of the intelligent agent; 3, constructing a roll phase plane switching module based on the roll angle and the roll angle speed, and designing a switching boundary and a preemption and delay switching mechanism of the roll phase plane switching module; 4, training the SAC intelligent agent to obtain an optimal vertical-roll cooperative control strategy; and 5, inputting the current state of the vehicle into the roll phase plane switching module and the SAC intelligent agent, determining the current control mode, and independently outputting the left side active force and the right side active force to act on the vehicle. According to the method, collaborative optimization of vehicle roll stability and vehicle body smoothness can be achieved under the composite working condition of bilateral asymmetric excitation steering driving, and working condition change adaptability is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle active suspension control, and particularly relates to a vertical-roll cooperative control method based on roll phase plane switching, which is especially applicable to the comprehensive optimization of vehicle stability and comfort under the combined working conditions of bilateral asymmetric excitation road surface and steering. Background Art

[0002] In the process of vehicle intelligent development, multi-objective cooperative control under complex working conditions has become a key technical challenge. When a vehicle steers on an asymmetric excitation road surface with significant differences on both sides, there is a high degree of coupling between the vertical vibration and roll motion of the vehicle body. Traditional active suspension control strategies often adopt hierarchical design or fixed parameter models, which are difficult to solve the conflict between the objectives of vertical comfort and roll stability. In addition, the mode switching strategy based on threshold triggering is prone to cause control command oscillation due to road surface mutation, resulting in energy waste of the suspension actuator and performance imbalance. Therefore, there is an urgent need for an active suspension control method that integrates real-time state perception and multi-modal cooperative optimization to break through the performance bottleneck under dynamic coupling working conditions. Summary of the Invention

[0003] In order to overcome the deficiencies of the prior art, the present invention provides a vertical-roll cooperative control method for vehicles based on roll phase plane switching, aiming to realize the cooperative optimization of roll stability and vertical comfort under the combined working conditions of steering on a bilateral asymmetric road surface excitation, broaden the vehicle roll stability domain and improve the riding comfort.

[0004] In order to achieve the above object of the invention, the following technical solutions are adopted:

[0005] A vertical-roll cooperative control method for vehicles based on roll phase plane switching according to the present invention is characterized in that it is applied to an active suspension vehicle and is carried out according to the following steps:

[0006] Step 1: Based on the Dugoff tire model, establish a six-degree-of-freedom steering-roll vehicle dynamics model to describe the vertical vibration of the sprung and unsprung masses and the roll, yaw, and lateral motion of the vehicle;

[0007] Step 2: Construct an intelligent agent based on the soft actor-critic algorithm, and design the state space, action space, multi-mode reward function, and network structure of the intelligent agent;

[0008] Step 3: Based on the roll angle and roll angular velocity, construct a roll phase plane switching module, and design its mode switching boundary and preemption and delay mechanisms;

[0009] Step 4: Under the input of left and right asymmetric random road surface excitation and random front wheel steering angle, use the roll phase plane switching module to train the intelligent agent until the total reward converges to a high-level reward value, so as to obtain the optimal intelligent agent for offline control;

[0010] Step 5: Input the vehicle body roll angle and angular velocity into the roll phase plane switching module to determine the current control mode, so that the optimal agent outputs the active forces on the left and right suspensions in the current control mode to achieve the coordinated optimization control of vertical comfort and roll stability.

[0011] Another feature of the vehicle vertical-roll coordinated control method based on roll phase plane switching according to the present invention is that the step 2 includes:

[0012] Step 2.1: Define the state space , where is the vehicle lateral acceleration, is the front wheel steering angle, is the vehicle body roll angle, is the vehicle body roll angular velocity, is the roll angle tracking error, and are the displacements of the unsprung masses on the left and right sides respectively, and are the velocities of the unsprung masses on the left and right sides respectively, is the vehicle body acceleration;

[0013] Step 2.2: Define the action space , , are the active forces output by the left and right active suspensions respectively;

[0014] Step 2.3: Construct different reward functions according to the mode :

[0015] Calculate the reward function in the vertical damping dominant mode CCM according to Equation (1) :

[0016] (1)

[0017] Calculate the reward function in the roll safety dominant mode SCM according to Equation (2) :

[0018] (2)

[0019] Calculate the additional penalty term according to Equation (3) :

[0020] (3)

[0021] In Equation (3), is the active force output limit value, and is the change value of the left and right active forces;

[0022] Calculate the total reward function according to Equation (4) :

[0023] (4)

[0024] Step 2.4: Design the network structure of the agent, including: Critic1 network, Critic2 network, target Critic1 network, target Critic2 network and Actor network; among them, the network structures of the Critic1 network, Critic2 network, target Critic1 network and target Critic2 network are exactly the same.

[0025] The Critic1 network, Critic2 network, target Critic1 network and target Critic2 network all include: a first fully connected layer, a feature splicing layer, a second fully connected layer, a third fully connected layer and an output layer connected in sequence; among them, the input of the feature splicing layer is the output of the first fully connected layer and the action, and the output layer generates the value of the state-action pair;

[0026] The Actor network includes: a fourth fully connected layer, a fifth fully connected layer, and an action output layer, and the action output layer includes: a mean branch, a variance branch and a Gaussian distribution sampling layer;

[0027] Among them, the mean branch includes: a first sub-fully connected layer and a mean output layer connected in sequence, the input end of the first sub-fully connected layer is connected to the output end of the fifth fully connected layer, and the mean output layer is used to output the first vector of the active force mean value;

[0028] The variance branch includes: a second sub-fully connected layer and a variance output layer connected in sequence, the input end of the second sub-fully connected layer is connected to the output end of the fifth fully connected layer, and the variance output layer is used to output the second vector of the active force variance;

[0029] The Gaussian distribution sampling layer constructs a Gaussian distribution using the first vector and the second vector of the active force, and generates the active forces of the left and right active suspensions.

[0030] Furthermore, the step 3 includes:

[0031] Step 3.1: Construct a roll phase plane:

[0032] Input the sine front wheel angle excitation into the six-degree-of-freedom steering-roll vehicle dynamics model or the real vehicle to obtain the roll angle , roll angle velocity and the lateral load transfer rate LTR, with |LTR| > 0.9 as the instability boundary, plot phase plane;

[0033] Step 3.2: Determine the mode switching boundary:

[0034] Define the roll stability factor , where and are two coefficient factors, is the set switching limit value. When exceeds the set value , the vehicle directly enters the roll safety dominant mode SCM; otherwise, the vehicle will initiate a demand to enter the vibration damping dominant mode CCM and perform the actual mode switching according to the preemption and delay mechanism shown in Equation (5);

[0035] (5)

[0036] In Equation (5), represents the control mode at time represents the control mode at time is the time of the last entry into SCM and is given by Equation (6), is the control step size, and N is the number of control cycles for the set delay switching;

[0037] (6).

[0038] Furthermore, the said Step 4 includes:

[0039] Step 4.1: Randomly initialize the parameters of the Critic1 network and the parameters of the Critic2 network ; and assign the parameters of the Critic1 network to the parameters of the target Critic1 network and assign the parameters of the Critic2 network to the parameters of the target Critic2 network ; randomly initialize the parameters of the Actor network ;

[0040] Step 4.2: Input the state at time into the Actor network, output the action at time and apply it to the vehicle to obtain the state at time ; use Calculate the roll factor based on the roll angle and angular velocity in , and determine it by the preemption and delay mechanism shown in Equation (5) The control mode at the moment, and then calculate The reward at the moment , so as to obtain a sample data And store it in the experience pool;

[0041] Step 4.3: Randomly extract a sample data from the experience pool , and And Input into the Critic1 network and the Critic2 network, and correspondingly output the first value of the state-action pair And the second value ;

[0042] Input Into the Actor network, and output The action at the moment ;

[0043] Input And Into the target Critic1 and target Critic2 networks, and correspondingly output the first target value of the state-action pair And the second target value ;

[0044] Calculate according to Equation (7) The target value at the moment :

[0045] (7)

[0046] In Equation (7), Is the discount factor;

[0047] Step 4.4: Respectively With And Do mean square error to construct the Loss function, so as to use the gradient descent method to update the parameters of the Critic1 network , the parameters of the Critic2 network , and get the updated parameters of the Critic1 network and assign them to , the updated parameters of the Critic2 network and assign them to ;

[0048] Step 4.5: Use Equation (8) to update the parameters of the target Critic1 network And the parameters of the target Critic2 network :

[0049] (8)

[0050] In formula (8), is the smoothing factor, is the assignment symbol;

[0051] Step 4.6: According to , adopt a delayed update mechanism to update the parameters of the Actor network , obtain the updated parameters of the Actor network and assign them to , the delay mechanism means that the Actor network is updated after the Critic network is updated several times;

[0052] Step 4.7: Iteratively train the agent according to the process of Step 4.2 - Step 4.6 until the reward converges to a stable high reward, so as to obtain the optimal agent for offline deployment.

[0053] Furthermore, the roll angle tracking error is calculated as follows:

[0054] Step 5.1: Construct the ideal body reverse roll angle according to formula (9):

[0055] (9)

[0056] In formula (9), is the longitudinal vehicle speed, , are the distances from the vehicle's center of mass to the front axle and the rear axle respectively, is the gravitational acceleration.

[0057] Step 5.2: Construct the target body reverse roll angle according to formula (10):

[0058] (10)

[0059] In formula (10), is the proportionality factor, is the set reverse roll angle limit;

[0060] Step 5.3: Construct the roll angle tracking error according to formula (11):

[0061] (11)

[0062] An electronic device of the present invention includes a memory and a processor, characterized in that the memory stores a program for implementing the active suspension reverse roll control method, and when the processor executes the program, active suspension reverse roll control is achieved.

[0063] A computer-readable storage medium of the present invention stores a computer program, characterized in that when the computer program is executed by a processor, the steps of the active suspension reverse roll control method are achieved.

[0064] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0065] 1. By deeply integrating the roll phase plane switching mechanism with the vertical comfort-dominated mode and the roll safety-dominated mode into the same SAC intelligent agent framework, the present invention realizes the dynamic distribution of suspension active force and multi-objective collaborative optimization under the combined working conditions of bilateral asymmetric excitation road surface and steering driving; at the same time, by utilizing the continuous action space characteristics of SAC, the problem of sudden change of control force caused by traditional switch-type switching is eliminated, and the consistency guarantee ability of vehicle roll stability and vertical comfort is significantly improved.

[0066] 2. By designing a mode preemption trigger and delay exit mechanism, in the scenarios of high-speed emergency obstacle avoidance or continuous roll disturbance, the execution priority of the roll safety-dominated mode is preferentially locked, the unnecessary activation of the vertical comfort-dominated mode is suppressed, the oscillation of the suspension actuator caused by high-frequency switching is avoided, and the persistence of roll stability intervention is ensured by delaying the minimum dwell time constraint of multiple control cycles, and finally the global enhanced control of roll safety priority is realized. Description of the Drawings

[0067] Figure 1 It is the overall control strategy diagram of the present invention;

[0068] Figure 2 It is the structure diagram of the Critic network of the SAC intelligent agent of the present invention;

[0069] Figure 3 It is the structure diagram of the Actor network of the SAC intelligent agent of the present invention;

[0070] Figure 4 It is the schematic diagram of the roll phase plane of the present invention;

[0071] Figure 5 It is the schematic diagram of the mode switching boundary of the present invention;

[0072] Figure 6 It is the execution flow of the SAC algorithm of the present invention. Detailed Embodiments

[0073] In this embodiment, a vehicle vertical-roll coordinated control method based on roll phase-plane switching is applied to a vehicle with an active suspension. The overall control block diagram is as shown in Figure 1 and is carried out according to the following steps:

[0074] Step 1: Establish a six-degree-of-freedom steering-roll vehicle dynamics model to describe the vertical vibrations of the sprung and unsprung masses and the roll, yaw, and lateral motions of the vehicle. Select the Dugoff model to establish the coupling relationship between the vertical and lateral forces of the tire.

[0075] Step 1.1: Use Equation (1) to establish a six-degree-of-freedom vehicle dynamics model considering lateral, yaw, roll, and vertical vibrations:

[0076] (1)

[0077] In Equation (12), is the body mass, is the total vehicle mass, , are the unsprung masses on the left and right sides respectively, , are the vehicle yaw moment of inertia and the body roll moment of inertia respectively, , are the passive vertical forces of the left and right suspensions respectively, , are the active forces of the left and right suspensions respectively, , are the lateral forces of the front and rear wheels respectively, is the tire stiffness, , , , , are the body displacement, the displacement of the unsprung mass on the left side, the displacement of the unsprung mass on the right side, the left road excitation input, and the right road excitation input respectively, , , are the body acceleration, the acceleration of the unsprung mass on the left side, and the acceleration of the unsprung mass on the right side respectively, is the vehicle yaw angle, is the vehicle yaw angular acceleration, , are the distances from the center of mass to the front and rear axles respectively, is the vehicle lateral acceleration, calculated according to Equation (2), is the body roll angle, is the body roll angular acceleration, is the half track width, is the distance from the roll center to the vehicle center of mass, is the active roll moment.

[0078] (2)

[0079] In Equation (2), is the lateral displacement of the vehicle, is the longitudinal vehicle speed.

[0080] Step 1.2: Calculate the passive vertical forces of the suspension stiffness and damping on both sides according to Equation (3):

[0081] (3)

[0082] In Equation (3), is the stiffness coefficient of the suspension spring, is the damping coefficient of the suspension shock absorber, , are the displacements of the unsprung masses on the left and right sides respectively, , are the velocities of the unsprung masses on the left and right sides respectively, , are the velocities of the sprung masses on the left and right sides respectively.

[0083] Step 1.3: Based on the small roll angle assumption, there is , and the relationship between the motions of the sprung masses on the left and right sides and the motions of the unsprung masses is expressed according to Equation (4):

[0084] (4)

[0085] Step 1.4: Based on the small front wheel steering angle assumption, the front and rear wheel sideslip angles are calculated according to the bicycle steering model, as shown in Equation (5):

[0086] (5)

[0087] In Equation (5), , are the front and rear wheel tire sideslip angles respectively, is the front wheel steering angle.

[0088] Step 1.5: Select the Dugoff tire model to characterize the coupling relationship between the tire lateral force and the vertical load, and calculate the tire lateral force according to Equation (6):

[0089] (6)

[0090] In Equation (6), is the road surface adhesion coefficient, is the tire slip ratio, is the tire sideslip angle, is the tire longitudinal stiffness, is the lateral stiffness of the tire, is the speed factor, is the vertical force of the tire.

[0091] Step 2: Based on the reverse roll mechanism, generate a gravitational moment component opposite to the centrifugal moment by actively tilting the vehicle body to improve the vehicle roll stability and lateral comfort, and calculate the target vehicle roll angle , and then calculate the roll angle tracking error .

[0092] Step 2.1: Calculate the centrifugal moment according to Equation (7) and the gravitational moment

[0093] (7)

[0094] When the vehicle is driving steadily, , calculate the ideal yaw acceleration response when the vehicle is steering according to Equation (8):

[0095] (8)

[0096] Step 2.3: Let = , and combine with Equation (8) to obtain the calculation of the ideal reverse roll angle as Equation (9):

[0097] (9)

[0098] Considering the output limit of the active suspension, power consumption optimization, and the requirement for rapid adjustment of the roll angle, it is not necessary to track the ideal roll angle. The actual tracked target roll angle is calculated according to Equation (10):

[0099] (10)

[0100] In the formula, is the proportionality factor, is the set limit value of the active roll angle.

[0101] Step 2.5: Construct the roll angle tracking error according to Equation (11):

[0102] (11)

[0103] Step 3: Construct an agent based on the Soft Actor-Critic (SAC) algorithm, and design the state space, action space, multi-mode reward function, and network structure of the SAC agent.

[0104] Step 3.1: Define the state space: When steering on an asymmetric road surface with significant height fluctuations, there is a coupling between the vertical vibration and roll motion of the vehicle body. The selected state space should be able to completely describe the characteristics of vertical vibration and roll motion. At the same time, the selected observed states should be closely related to the optimization objective. Selecting too many states that are less related to the optimization objective will have a negative impact on the learning of the SAC agent. Therefore, select ;

[0105] Step 3.2: Define the action space: Independent output of the left and right active forces is required to meet the collaborative optimization requirements of vertical and roll under road surface excitation. Therefore, select ;

[0106] Step 3.3: Define the reward function: When the vehicle is driving on a fluctuating road surface, it is not necessary to strictly track the ideal roll angle when the steering angle is small, which will instead lead to deterioration of the vehicle body vibration; when the steering angle is large, it is necessary to strictly track the ideal roll angle to expand the lateral safety domain; under large rough asymmetric excitations on both sides, roll instability may also occur, and at this time, the suppression of the roll angle becomes important again. Therefore, when jointly controlling vibration and roll, it is also necessary to dynamically divide the weights of the two. For this, reward functions with different optimization focuses are designed for the two control modes. Divided by control mode into:

[0107] Vertical vibration reduction dominant mode (CCM), the core is to reduce the vehicle body acceleration response without requiring precise tracking of the ideal roll angle, and the reward is calculated according to Equation (12):

[0108] (12)

[0109] Roll safety dominant mode (SCM), the core is to track the dynamic target roll angle to ensure lateral stability without excessive optimization of vertical vibration control, and the reward is calculated according to Equation (13):

[0110] (13)

[0111] To ensure the continuous change of the active force and reduce the actuator energy consumption, an additional penalty term is designed and calculated according to Equation (14):

[0112] (14)

[0113] In Equation (14), is the limit value of the suspension active force, which is 6000N in this example, and are the change values of the left and right active forces.

[0114] The total reward function is calculated according to Equation (15):

[0115] (15)

[0116] Step 3.4: Design the network structure of the SAC agent, including: Critic1, Critic2 networks and target Critic1, Critic1 networks. The structures of the above four networks are the same and are collectively referred to as the Critic network; an Actor network, a total of 5 networks;

[0117] Each Critic network includes: a first fully connected layer, a feature splicing layer, a second fully connected layer, a third fully connected layer, and an output layer connected in sequence; among them, the input of the feature splicing layer is the output of the first fully connected layer and the action, and the output layer generates the value of the state-action pair;

[0118] The Actor network includes: a fourth fully connected layer, a fifth fully connected layer, and an action parameter output layer. The action parameter output layer includes:

[0119] A mean branch, including a first sub-fully connected layer and a mean output layer connected in sequence. The input end of the first sub-fully connected layer is connected to the output end of the fifth fully connected layer, and the mean output layer is used to output the first vector of the mean of the active force;

[0120] A variance branch, including a second sub-fully connected layer and a variance output layer connected in sequence. The input end of the second sub-fully connected layer is connected to the output end of the fifth fully connected layer, and the variance output layer is used to output the second vector of the variance of the active force;

[0121] A Gaussian distribution sampling layer, which constructs a Gaussian distribution using the first vector and the second vector of the active force and generates the active forces of the left and right active suspensions.

[0122] In this example, the structures of the Critic and Actor networks are shown in Figure 2 and Figure 3 . The input of the Critic network is 10 state variables and 2 action variables. After the state variables are dimensionally enhanced by a fully connected layer with 64 nodes, they are combined with 2 action variables, and then the Q value is output through two hidden layers with 512 nodes and 256 nodes. The activation function is selected as Leaky ReLU. The input of the Actor network is 10 state variables. After passing through two hidden layers with 512 and 256 nodes respectively, two branches are led out, and the mean and standard deviation of the action are output through hidden layers with 32 nodes. The activation function is also selected as Leaky ReLU to reduce the occurrence of the "necrosis" phenomenon of neurons. Both the Actor and Critic networks use the adam optimizer. The Critic learning rate is 5×10 -4 , and the Actor learning rate is 1×10 -4 .

[0123] Step 4: Construct a roll phase plane switching module, and design its mode switching boundary, preemption, and delay mechanisms.

[0124] Step 4.1: Roll phase plane construction: By inputting sinusoidal front wheel angles with different amplitudes and frequencies into a steering-roll vehicle model or a real vehicle, the vehicle roll angle and roll angle speed responses are obtained, and the corresponding lateral load transfer ratio LTR is calculated. The change range of the front wheel angle amplitude is 0 to 20 degrees, and the frequency change range is 2 to 10 Hz, that is, the fastest front wheel angle rate can reach 40 deg / s, which fully covers the steering speed of the steering wheel under the extreme driving conditions of the driver. Using the three-dimensional data obtained by simulation, taking as the vehicle instability, the roll phase plane can be drawn, as shown in Figure 4 .

[0125] Step 4.2: Mode switching threshold: In roll stability control, if the control is only intervened at the instability boundary, it is often impossible to quickly correct the attitude because the system already has a large inertia at this time. At the same time, in this embodiment, the SCM and CCM also require appropriate switching boundaries, so the roll stability factor is selected to define the mode switching boundary. In the formula, , are two coefficient factors, is the set switching limit value. When exceeds the set value , it immediately enters the SCM mode; otherwise, the vehicle will initiate a demand to enter the damping-dominant mode CCM and perform the actual mode switching according to the delay mechanism in Step 3.3. In this example, , , , and the schematic diagram of the mode switching boundary is obtained as shown in Figure 5 .

[0126] Step 4.3: SCM preemption and delay mechanism: The SCM mode can preempt the CCM mode. After entering the SCM mode, it is necessary to maintain control cycles before the mode can be switched. This mechanism can avoid high-frequency switching of the mode when the vehicle state fluctuates greatly and ensure the priority of vehicle safety control.

[0127] The mode switching logic is calculated according to Equation (16):

[0128] (16)

[0129] In Equation (16), represents the control mode at time, represents Control mode at a moment is the time when entering the SCM mode last time is the system control step size

[0130] The timestamp update rule is shown in Equation (17):

[0131] (17)

[0132] Step 5: The training of the SAC agent is carried out as follows. The SAC training execution framework is shown in Figure 6 ;

[0133] Step 5.1: Randomly initialize the parameters of the Critic1 network and the parameters of the Critic2 network ; and assign the parameters of the Critic1 network to the parameters of the target Critic1 network and assign the parameters of the Critic2 network to the parameters of the target Critic2 network ; randomly initialize the parameters of the Actor network ;

[0134] Step 5.2: Input the state at the moment into the Actor network, output the action at the moment and apply it to the vehicle to obtain the state at the moment ; calculate the roll factor using the roll angle and angular velocity in , and determine the control mode at the moment by the preemption and delay mechanism of Equation (5), and then calculate the reward at the moment , so as to obtain a sample data and store it in the experience pool

[0135] Step 5.3: Randomly extract a sample data from the experience pool, input and into the Critic1 network and the Critic2 network, and respectively output the first value and the second value of the state-action pair;

[0136] Input into the Actor network and output the action at the moment ;

[0137] Input and into the target Critic1 and Critic2 networks, and correspondingly output the first target value of the value of the state-action pair and the second target value ;

[0138] Calculate according to Equation (18) the target value at time :

[0139] (18)

[0140] In Equation (18), is the discount factor, which is taken as 0.99 in this example.

[0141] Step 5.4: Take respectively and and to calculate the mean square error to construct the Loss function, and thus use the gradient descent method to update the parameters of the Critic1 network, the parameters of the Critic2 network are updated, and the updated parameters of the Critic1 network are assigned to .

[0142] Step 5.5: Update the parameters of the target Critic1 network and the parameters of the target Critic2 network

[0143] (19)

[0144] In Equation (19), is the smoothing factor, which is taken as 0.001 in this example, is the assignment symbol.

[0145] Step 5.6: According to , adopt the delayed update mechanism to update the parameters of the Actor network, and the updated parameters of the Actor network are assigned to .

[0146] Step 5.7: Iteratively train the agent according to the process of Step 5.2 - Step 5.6 until the reward converges to a stable high reward, so as to obtain the optimal agent for offline deployment.

[0147] Step 6: Input the current vehicle state into the roll phase plane switching module and the SAC agent, determine the current control mode, and independently output the left and right active forces to achieve the comprehensive optimization of roll stability and ride comfort.

[0148] In this embodiment, an electronic device includes a memory and a processor. The memory is used to store a program that supports the processor to execute the above method, and the processor is configured to execute the program stored in the memory.

[0149] In this embodiment, a computer-readable storage medium stores a computer program, and when the computer program is run by a processor, it executes the steps of the above method.

Claims

1. A vehicle vertical-roll cooperative control method based on roll phase plane switching, characterized in that It is applied to an active suspension vehicle and is carried out according to the following steps: Step 1: Based on the Doguff tire model, establish a six-degree-of-freedom steering-roll vehicle dynamics model to describe the vertical vibrations of the sprung and unsprung masses and the roll, yaw, and lateral motions of the vehicle; Step 2: Construct an agent based on the soft actor-critic algorithm and design the state space, action space, multi-mode reward function, and network structure of the agent; Step 3: Based on the roll angle and roll angle velocity, construct a roll phase plane switching module and design its mode switching boundary and preemption and delay mechanisms; Step 4: Under the excitation of left and right asymmetric random road surfaces and random front wheel angle inputs, use the roll phase plane switching module to train the agent until the total reward converges to a high-level reward value, thereby obtaining an optimal agent for offline control; Step 5: Input the body roll angle and angular velocity into the roll phase plane switching module to determine the current control mode, so that the optimal agent outputs the active forces on the left and right suspensions in the current control mode according to the current vehicle state, to achieve the coordinated optimization control of vertical comfort and roll stability.

2. The vehicle vertical-roll cooperative control method based on roll phase plane switching according to claim 1, characterized in that, The said Step 2 includes: Step 2.1: Define the state space , where is the vehicle lateral acceleration, is the front wheel steering angle, is the body roll angle, is the body roll rate, is the roll angle tracking error, and are the wheel unsprung mass displacements on the left and right sides, respectively, and are the wheel unsprung mass velocities on the left and right sides, respectively, is the body acceleration; Step 2.2: Define the action space , and are the active forces output by the left and right active suspensions respectively; Step 2.3: According to the mode Construct different reward functions: Calculate the reward function in the dominant vertical vibration damping mode CCM according to Equation (1). : (1) Calculate the reward function under the roll safety dominant mode SCM according to Equation (2). : (2) Calculate the additional penalty term according to Equation (3). : (3) In formula (3), is the limit value of the active power output, and are the change values of the active power on the left and right sides; Calculate the total reward function according to Equation (4). :[[]]END]] (4) Step 2.4: Design the network structure of the agent, including: Critic1 network, Critic2 network, target Critic1 network, target Critic2 network, and Actor network; among them, the network structures of the Critic1 network, Critic2 network, target Critic1 network, and target Critic2 network are all exactly the same; The Critic1 network, Critic2 network, target Critic1 network, and target Critic2 network all include: a first fully connected layer, a feature concatenation layer, a second fully connected layer, a third fully connected layer, and an output layer connected in sequence; among them, the input of the feature concatenation layer is the output of the first fully connected layer and the action, and the output layer generates the value of the state-action pair; The Actor network includes: a fourth fully connected layer, a fifth fully connected layer, and an action output layer, and the action output layer includes: a mean branch, a variance branch, and a Gaussian distribution sampling layer; Among them, the mean branch includes: a first sub-fully connected layer and a mean output layer connected in sequence, the input end of the first sub-fully connected layer is connected to the output end of the fifth fully connected layer, and the mean output layer is used to output the first vector of the active force mean; The variance branch includes: a second sub-fully connected layer and a variance output layer connected in sequence, the input end of the second sub-fully connected layer is connected to the output end of the fifth fully connected layer, and the variance output layer is used to output the second vector of the active force variance; The Gaussian distribution sampling layer constructs a Gaussian distribution using the first vector and the second vector of the active force and generates the active forces of the left and right active suspensions.

3. A vehicle vertical-roll cooperative control method based on roll phase plane switching according to claim 1, characterized in that, The said Step 3 includes: Step 3.1: Construct a roll phase plane: Input the sine front wheel steering angle excitation into a six-degree-of-freedom steering-roll vehicle dynamics model or a real vehicle to obtain the roll angle , the roll angle velocity and the lateral load transfer ratio LTR. Taking |LTR|>0.9 as the instability boundary, plot the phase plane; Step 3.2: Determine the mode switching boundary: Define the roll stability factor , where 、 are two coefficient factors is the set switching limit value. When exceeds the set value , the vehicle directly enters the roll safety dominant mode SCM; otherwise, the vehicle will initiate a demand to enter the damping dominant mode CCM and perform the actual mode switching according to the preemption and delay mechanism shown in Equation (5). (5) In formula (5), represents the control mode at time represents the control mode at time is the time of the last entry into the SCM, and from formula (6), is the control step size, and N is the number of control cycles for the set delay switching; (6)。 4. A vehicle vertical-roll cooperative control method based on roll phase-plane switching according to claim 1, characterized in that The said Step 4 includes: Step 4.1: Randomly initialize the parameters of the Critic1 network , the parameters of the Critic2 network ; and assign the parameters of the Critic1 network to the parameters of the target Critic1 network , assign the parameters of the Critic2 network to the parameters of the target Critic2 network ; Randomly initialize the parameters of the Actor network ; Step 4.2: Input the state at moment into the Actor network, and output the action at moment, and apply it to the vehicle to obtain the state at moment; Calculate the roll factor using the roll angle and angular velocity in , and determine the control mode at moment by the preemption and delay mechanism shown in Equation (5), and then calculate the reward at moment, so as to obtain a sample data and store it in the experience pool; Step 4.3: Randomly draw a sample data from the experience pool , and input and into Critic1 network and Critic2 network, and correspondingly output the first value and the second value ; Input into the Actor network and output the action at time ; Input and into the target Critic1 and Critic2 networks, and correspondingly output the first target value of the value of the state-action pair and the second target value ; Calculate according to formula (7) The target value at the moment : (7) In formula (7), is the discount factor; Step 4.4: Respectively with and calculate the mean square error to construct the Loss function, and then use the gradient descent method to update the parameters of the Critic1 network, the parameters of the Critic2 network to obtain the updated parameters of the Critic1 network and assign them to ; Step 4.5: Use Equation (8) to update the parameters of the target Critic1 network and the parameters of the target Critic2 network : (8) In formula (8), is the smoothing factor, is the assignment symbol; Step 4.6: According to , the parameters of the Actor network are updated by adopting a delayed update mechanism , and the updated parameters of the Actor network are obtained and assigned to , where the delay mechanism means that the Actor network is updated after the Critic network has been updated several times; Step 4.7: Iteratively train the agent according to the process of Step 4.2 - Step 4.6 until the reward converges to a stable high reward, thereby obtaining an optimal agent for offline deployment.

5. A vehicle vertical-roll coordinated control method based on roll phase-plane switching according to claim 2, characterized in that The roll angle tracking error is calculated as follows: Step 5.1: Construct the ideal body reverse roll angle according to Equation (9) :[[]]END]] (9) In Equation (9), is the longitudinal speed of the vehicle, , are the distances from the vehicle's center of mass to the front axle and the rear axle respectively, is the acceleration due to gravity; Step 5.2: Construct the target body reverse roll angle according to Equation (10) :[[-END]] (10) In formula (10), is the scale factor, is the set reverse roll angle limit value; Step 5.3: Construct the roll angle tracking error according to Equation (11) :[[]]END]] (11)。 6. An electronic device, comprising a memory and a processor, characterized in that, The memory stores a program for implementing the vehicle vertical-roll cooperative control method described in any one of claims 1-5, and when the processor executes the program, active suspension reverse roll control is implemented.

7. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, the steps of the vehicle vertical-roll cooperative control method described in any one of claims 1-5 are implemented.

Images