A vehicle body stability control method and system

Abstract

The application provides a vehicle body stability control method and system, which comprises the following steps: configuring a braking system, a steering system and a suspension system in a vehicle interior as independent pre-control agents respectively, training each pre-control agent into a corresponding cooperative pre-control model through a multi-agent learning algorithm based on the sensing characteristics and control response parameters of each system; collecting corresponding target data through each pre-control agent, and judging whether the vehicle has a stability loss risk according to the target data through the cooperative pre-control model; if it is judged that the vehicle has a stability loss risk according to the target data through the cooperative pre-control model, outputting a multi-system cooperative pre-control instruction adapted to each pre-control agent in advance by a preset time; and driving the corresponding system to perform a pre-control action through the multi-system cooperative pre-control instruction, so that the vehicle is in a stable driving state. The application can avoid the occurrence of the stability loss phenomenon of the vehicle body, and correspondingly improves the vehicle body control efficiency.

Description

Technical Field

[0001] This invention relates to the field of automotive technology, and in particular to a vehicle stability control method and system. Background Technology

[0002] With the advancement of technology and the rapid development of productivity, automobiles have become widespread in people's daily lives and have become an indispensable means of transportation, greatly facilitating people's lives.

[0003] To improve vehicle safety, existing car manufacturers have developed corresponding vehicle stability systems, which are widely used to implement preventative measures when a car may tilt, skid, or lose its center of gravity, in order to maintain a safe driving state.

[0004] Furthermore, in practical applications, most existing vehicle stability control systems rely on actual feedback from sensors, which is a reactive mode. They often intervene only when the vehicle has already shown a clear tendency to become unstable, resulting in a certain degree of lag and reducing the efficiency of vehicle stability. Summary of the Invention

[0005] Based on this, the purpose of the present invention is to provide a vehicle stability control method and system to solve the problem that in the process of controlling vehicle stability in the prior art, intervention is mostly only carried out when the vehicle has already shown obvious signs of instability, resulting in a certain degree of lag.

[0006] The first aspect of the present invention proposes:

[0007] A vehicle stability control method specifically includes the following steps:

[0008] The braking system, steering system, and suspension system inside the vehicle are each configured as an independent pre-control agent. Based on the perception characteristics and control response parameters of each system, the pre-control agents are trained into corresponding collaborative pre-control models through a multi-agent learning algorithm.

[0009] Each of the aforementioned pre-control intelligent agents collects corresponding target data, and the collaborative pre-control model determines whether the vehicle has any potential instability risks based on the target data.

[0010] If the collaborative pre-control model determines that the vehicle has a potential for instability based on the target data, it will output a multi-system collaborative pre-control command adapted to each of the pre-control agents at a preset time.

[0011] The multi-system collaborative pre-control command drives the corresponding system to perform pre-control actions, so as to keep the vehicle in a stable driving state.

[0012] The beneficial effects of this invention are as follows: by configuring the systems inside the vehicle as corresponding pre-control agents, a collaborative pre-control model for subsequent analysis can be trained. Based on this, each pre-control agent collects corresponding target data, and the collaborative pre-control model determines whether the vehicle will experience instability. If so, an appropriate multi-system collaborative pre-control command can be generated immediately, and the corresponding execution action can be completed in advance, thereby effectively preventing vehicle instability and improving the vehicle's control efficiency.

[0013] Furthermore, the step of training each of the pre-control agents into corresponding collaborative pre-control models using a multi-agent learning algorithm based on the perception characteristics and control response parameters of each system includes:

[0014] The braking force output range of the braking system, the steering angle adjustment accuracy of the steering system, and the stiffness response rate of the suspension system are collected respectively, and the corresponding characteristic parameter matrix is ​​established by combining the signal transmission delay of the system sensors.

[0015] The characteristic parameter matrix is ​​fused with vehicle instability sample data under different road conditions, and an independent training network is assigned to each pre-control agent. At the same time, training convergence conditions are set to complete the initial training of each pre-control agent.

[0016] A multi-agent cooperative communication protocol is constructed. Based on the model after initial training, the timing synchronization error and amplitude conflict of the control commands of each pre-control agent are optimized through multiple rounds of iteration to generate the cooperative pre-control model.

[0017] Furthermore, the step of generating the collaborative pre-control model by iteratively optimizing the time synchronization error and amplitude conflict of the control commands of each pre-control agent based on the initialized trained model includes:

[0018] Based on the characteristic parameter matrix, dynamic weight coefficients are constructed, and the control priority of each pre-control agent is allocated under each driving condition of the vehicle through the dynamic weight coefficients.

[0019] During each iteration, the timing of the instruction output of each pre-control agent is adjusted according to the control priority, and the superposition deviation of the control amplitude between different pre-control agents is corrected.

[0020] After each round of iterative optimization, the optimized model is connected to the vehicle hardware-in-the-loop simulation platform and verified based on a preset instability test set until the verification requirements are met, so as to generate the collaborative pre-control model.

[0021] Furthermore, the step of determining whether the vehicle has a potential for instability based on the target data using the collaborative pre-control model includes:

[0022] A scene weight matrix is ​​constructed based on the vehicle's driving speed and road surface adhesion coefficient, and the target data is then fused using the scene weight matrix in a scene-based weighted manner to generate a unified data sample.

[0023] The unified data samples are subjected to time-series correlation analysis to extract the instability characteristic parameters of the vehicle.

[0024] The instability characteristic parameters are input into the dynamic decision layer of the collaborative pre-control model to determine whether the vehicle has potential instability based on the instability judgment conditions.

[0025] Furthermore, the step of inputting the instability characteristic parameters into the dynamic decision layer of the collaborative pre-control model to determine whether the vehicle has potential instability based on the instability judgment conditions includes:

[0026] Based on the differences in the control response parameters of each of the pre-controlled intelligent agents, the instability characteristic parameters are prioritized to generate a corresponding target sequence list;

[0027] According to the target sequence list, each instability feature parameter is sequentially input into the corresponding hierarchical decision node of the dynamic decision layer, and the instability feature parameter received by each hierarchical decision node is subjected to temporal dimension enhancement processing to generate the corresponding temporal enhanced feature sequence.

[0028] The instability trend value corresponding to the vehicle is calculated based on the time-series enhanced feature sequence, and the magnitude of the instability trend value is used to determine whether the vehicle has potential instability risks.

[0029] Furthermore, the step of outputting multi-system collaborative pre-control instructions adapted to each of the pre-controlled intelligent agents at a preset time includes:

[0030] The vehicle's driving condition data is collected, and a dynamic time algorithm is used to calculate the advance time adapted to the vehicle based on the driving condition data.

[0031] Based on the data acquisition deviation value of each pre-control agent and combined with the training parameters of the collaborative pre-control model, the pre-control command parameter value corresponding to each pre-control agent is determined.

[0032] The multi-system collaborative pre-control command is generated based on the advance time and the pre-control command parameter value.

[0033] Furthermore, the step of generating the multi-system collaborative pre-control command based on the advance time and the pre-control command parameter value includes:

[0034] Based on the pre-control execution characteristics of each pre-control agent, the pre-control instruction parameter values ​​are divided into several independent parameter subsets;

[0035] Based on the advance time, and combined with the control response delay of each of the pre-control agents, a corresponding instruction triggering sequence is assigned to each of the independent parameter subsets;

[0036] Each independent parameter subset and its corresponding instruction trigger timing are associated and integrated to generate the multi-system collaborative pre-control instruction.

[0037] The second aspect of the present invention proposes:

[0038] A vehicle stability control system, wherein the system includes:

[0039] The training module is used to configure the braking system, steering system and suspension system inside the vehicle as independent pre-control agents, and to train each of the pre-control agents into a corresponding collaborative pre-control model based on the perception characteristics and control response parameters of each system through a multi-agent learning algorithm.

[0040] The data acquisition module is used to collect corresponding target data through each of the pre-control agents, and to determine whether the vehicle has any potential instability risks based on the target data through the collaborative pre-control model.

[0041] The output module is used to output multi-system collaborative pre-control instructions adapted to each of the pre-control agents at a preset time if the collaborative pre-control model determines that the vehicle has a risk of instability based on the target data.

[0042] The execution module is used to drive the corresponding system to perform pre-control actions through the multi-system collaborative pre-control instructions, so as to keep the vehicle in a stable driving state.

[0043] Furthermore, the training module is specifically used for:

[0044] The braking force output range of the braking system, the steering angle adjustment accuracy of the steering system, and the stiffness response rate of the suspension system are collected respectively, and the corresponding characteristic parameter matrix is ​​established by combining the signal transmission delay of the system sensors.

[0045] The characteristic parameter matrix is ​​fused with vehicle instability sample data under different road conditions, and an independent training network is assigned to each pre-control agent. At the same time, training convergence conditions are set to complete the initial training of each pre-control agent.

[0046] A multi-agent cooperative communication protocol is constructed. Based on the model after initial training, the timing synchronization error and amplitude conflict of the control commands of each pre-control agent are optimized through multiple rounds of iteration to generate the cooperative pre-control model.

[0047] Furthermore, the training module is specifically used for:

[0048] Based on the characteristic parameter matrix, dynamic weight coefficients are constructed, and the control priority of each pre-control agent is allocated under each driving condition of the vehicle through the dynamic weight coefficients.

[0049] During each iteration, the timing of the instruction output of each pre-control agent is adjusted according to the control priority, and the superposition deviation of the control amplitude between different pre-control agents is corrected.

[0050] After each round of iterative optimization, the optimized model is connected to the vehicle hardware-in-the-loop simulation platform and verified based on a preset instability test set until the verification requirements are met, so as to generate the collaborative pre-control model.

[0051] Furthermore, the acquisition module is specifically used for:

[0052] A scene weight matrix is ​​constructed based on the vehicle's driving speed and road surface adhesion coefficient, and the target data is then fused using the scene weight matrix in a scene-based weighted manner to generate a unified data sample.

[0053] The unified data samples are subjected to time-series correlation analysis to extract the instability characteristic parameters of the vehicle.

[0054] The instability characteristic parameters are input into the dynamic decision layer of the collaborative pre-control model to determine whether the vehicle has potential instability based on the instability judgment conditions.

[0055] Furthermore, the acquisition module is specifically used for:

[0056] Based on the differences in the control response parameters of each of the pre-controlled intelligent agents, the instability characteristic parameters are prioritized to generate a corresponding target sequence list;

[0057] According to the target sequence list, each instability feature parameter is sequentially input into the corresponding hierarchical decision node of the dynamic decision layer, and the instability feature parameter received by each hierarchical decision node is subjected to temporal dimension enhancement processing to generate the corresponding temporal enhanced feature sequence.

[0058] The instability trend value corresponding to the vehicle is calculated based on the time-series enhanced feature sequence, and the magnitude of the instability trend value is used to determine whether the vehicle has potential instability risks.

[0059] Furthermore, the output module is specifically used for:

[0060] The vehicle's driving condition data is collected, and a dynamic time algorithm is used to calculate the advance time adapted to the vehicle based on the driving condition data.

[0061] Based on the data acquisition deviation value of each pre-control agent and combined with the training parameters of the collaborative pre-control model, the pre-control command parameter value corresponding to each pre-control agent is determined.

[0062] The multi-system collaborative pre-control command is generated based on the advance time and the pre-control command parameter value.

[0063] Furthermore, the output module is specifically used for:

[0064] Based on the pre-control execution characteristics of each pre-control agent, the pre-control instruction parameter values ​​are divided into several independent parameter subsets;

[0065] Based on the advance time, and combined with the control response delay of each of the pre-control agents, a corresponding instruction triggering sequence is assigned to each of the independent parameter subsets;

[0066] Each independent parameter subset and its corresponding instruction trigger timing are associated and integrated to generate the multi-system collaborative pre-control instruction.

[0067] The third aspect of the present invention proposes:

[0068] A computer includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the vehicle stability control method as described above.

[0069] The fourth aspect of the present invention proposes:

[0070] A readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the vehicle stability control method as described above.

[0071] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0072] Figure 1 A flowchart of the vehicle stability control method provided in the first embodiment of the present invention;

[0073] Figure 2 This is a structural block diagram of the vehicle stability control system provided in the third embodiment of the present invention.

[0074] The following detailed description, in conjunction with the accompanying drawings, will further illustrate the present invention. Detailed Implementation

[0075] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Several embodiments of the invention are illustrated in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0076] It should be noted that when a component is said to be "fixed to" another component, it can be directly on the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0077] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0078] Please see Figure 1 The figure shows the vehicle stability control method provided in the first embodiment of the present invention. The vehicle stability control method provided in this embodiment can effectively avoid vehicle instability and improve the vehicle control efficiency accordingly.

[0079] Specifically, this embodiment provides:

[0080] A vehicle stability control method specifically includes the following steps:

[0081] Step S10: Configure the braking system, steering system and suspension system inside the vehicle as independent pre-control agents, and train each of the pre-control agents into a corresponding collaborative pre-control model based on the perception characteristics and control response parameters of each system through a multi-agent learning algorithm.

[0082] It's important to note that, firstly, the vehicle's braking system (responsible for deceleration and braking force distribution), steering system (responsible for direction adjustment), and suspension system (responsible for vehicle posture support) are configured as independent pre-control agents. Specifically, while their functions differ, they need to work together (e.g., during emergency steering, the braking system needs to assist in deceleration, and the suspension system needs to suppress body roll). This independent agent design fully leverages the perception and control characteristics of each system. Based on the perception characteristics of each system (e.g., the wheel speed perception accuracy of the braking system, the response speed of the steering angle sensor in the steering system) and control response parameters (e.g., brake pressure build-up time, steering motor response delay), a collaborative pre-control model is trained using a multi-agent learning algorithm (e.g., the MADDPG algorithm in reinforcement learning). This allows each agent to make autonomous judgments while also considering the actions of other systems, avoiding coordination conflicts caused by single-system control (e.g., mutual interference between over-braking and sharp steering). This facilitates subsequent processing.

[0083] Step S20: Collect corresponding target data through each of the pre-control intelligent agents, and determine whether the vehicle has potential instability risks based on the target data through the collaborative pre-control model;

[0084] It should be noted that, subsequently, each pre-control agent collects target data in real time: the braking system collects wheel speed and braking pressure; the steering system collects steering wheel angle and steering torque; and the suspension system collects vehicle roll angle and suspension travel. The collaborative pre-control model integrates and analyzes this data to determine whether the vehicle has potential instability risks (such as sideslip, fishtailing, understeer / oversteer) to facilitate subsequent processing.

[0085] Step S30: If the collaborative pre-control model determines that the vehicle has a potential for instability based on the target data, then the multi-system collaborative pre-control command adapted to each of the pre-control agents is output at a preset time in advance.

[0086] It's important to note that if instability is detected, the model will pre-set a timeframe (e.g., 0.5-1 second, dynamically adjusted based on vehicle speed and road conditions) to output multi-system collaborative pre-control instructions adapted to each agent. Specifically, this pre-control is crucial, allowing intervention before instability trends materialize, which is more efficient than traditional post-event correction. This facilitates subsequent processing.

[0087] Step S40: Drive the corresponding system to perform pre-control actions through the multi-system collaborative pre-control command so that the vehicle is in a stable driving state.

[0088] It should be noted that, ultimately, the command-driven system executes pre-control actions (such as differentiated braking in the braking system, minor steering corrections in the steering system, and stiffness adjustments in the suspension system) to collaboratively maintain vehicle stability and prevent instability, thus facilitating subsequent processing.

[0089] Second Embodiment

[0090] Furthermore, the step of training each of the pre-control agents into corresponding collaborative pre-control models using a multi-agent learning algorithm based on the perception characteristics and control response parameters of each system includes:

[0091] The braking force output range of the braking system, the steering angle adjustment accuracy of the steering system, and the stiffness response rate of the suspension system are collected respectively, and the corresponding characteristic parameter matrix is ​​established by combining the signal transmission delay of the system sensors.

[0092] The characteristic parameter matrix is ​​fused with vehicle instability sample data under different road conditions, and an independent training network is assigned to each pre-control agent. At the same time, training convergence conditions are set to complete the initial training of each pre-control agent.

[0093] A multi-agent cooperative communication protocol is constructed. Based on the model after initial training, the timing synchronization error and amplitude conflict of the control commands of each pre-control agent are optimized through multiple rounds of iteration to generate the cooperative pre-control model.

[0094] It's important to note that the first step is to establish a characteristic parameter matrix: This involves collecting data on the braking force output range of the braking system (e.g., 0-1500N), the steering angle adjustment accuracy of the steering system (e.g., ±0.1°), and the stiffness response rate of the suspension system (e.g., adjustment time from 0-1000N / mm), while also incorporating the signal transmission delays of each system's sensors (e.g., a 20ms delay for the brake pressure sensor). These parameters form the core elements of the matrix, quantifying the "capability boundaries" and "response speed" of each system, providing fundamental data for subsequent training.

[0095] The second step is model initialization training: The characteristic parameter matrix is ​​fused with vehicle instability sample data (such as wheel speed difference and vehicle roll angle during sideslip) under different road conditions (dry, wet, snowy) and driving conditions (rapid acceleration, sharp steering, emergency braking) to construct a training dataset. An independent training network is assigned to each pre-control agent (e.g., braking agent corresponds to the braking control network, steering agent corresponds to the steering control network), and training convergence conditions are set (e.g., the model's accuracy in predicting instability is ≥95%). Initialization training enables each agent to master the control logic of its own system under a single condition; for example, the braking agent learns that "a greater braking force should be applied to the inner wheel during sideslip."

[0096] The third step is collaborative optimization and model generation: A multi-agent collaborative communication protocol is constructed, defining the information interaction rules between agents (e.g., the braking agent sends the current braking torque to the steering agent, and the steering agent provides feedback on the expected steering angle). Based on the initialized and trained model, two core issues are addressed through multiple rounds of iterative optimization: first, the time synchronization error of control commands (e.g., the execution time difference between braking and steering actions must be ≤50ms to avoid misaligned intervention timing); second, control amplitude conflicts (e.g., the stiffness adjustment of the suspension system and the pressure output of the braking system may cause fluctuations in vehicle load, requiring correction of amplitude superposition deviations). After iteration, a collaborative pre-control model capable of seamless multi-system collaboration is generated for subsequent processing.

[0097] Furthermore, the step of generating the collaborative pre-control model by iteratively optimizing the time synchronization error and amplitude conflict of the control commands of each pre-control agent based on the initialized trained model includes:

[0098] Based on the characteristic parameter matrix, dynamic weight coefficients are constructed, and the control priority of each pre-control agent is allocated under each driving condition of the vehicle through the dynamic weight coefficients.

[0099] During each iteration, the timing of the instruction output of each pre-control agent is adjusted according to the control priority, and the superposition deviation of the control amplitude between different pre-control agents is corrected.

[0100] After each round of iterative optimization, the optimized model is connected to the vehicle hardware-in-the-loop simulation platform and verified based on a preset instability test set until the verification requirements are met, so as to generate the collaborative pre-control model.

[0101] It should be noted that, firstly, dynamic weighting coefficients are constructed based on the characteristic parameter matrix. These weighting coefficients reflect the control importance of each system under different driving conditions. Specifically, for example, the braking system has the highest weight when driving straight at high speed (responsible for stable deceleration), the steering system has the highest weight when making sharp turns at low speed (responsible for direction correction), and the suspension system has the highest weight on complex and bumpy roads (responsible for vehicle attitude control). These dynamic weighting coefficients assign clear control priorities to each pre-control agent, avoiding decision-making confusion.

[0102] In each iteration, the timing of command output is adjusted according to control priority: commands from high-priority systems are triggered first (e.g., steering commands are sent 20ms earlier than braking commands during sharp turns), and the delay of commands from low-priority systems is corrected using a timing compensation algorithm (e.g., suspension system commands are sent earlier according to their delay time). Simultaneously, deviations in the superposition of control amplitudes between different agents are corrected: for example, the braking force of the braking system increases the load on the front axle, which may cause changes in the steering torque demand of the steering system. The steering amplitude needs to be corrected according to the load transfer model to ensure vehicle stability after the two are superimposed.

[0103] After each iteration, the optimized model is connected to a vehicle hardware-in-the-loop (HiL) simulation platform. The platform simulates a pre-set set of instability test conditions (such as double lane change tests and braking tests on icy roads). Through interaction between the real ECU, sensors, and the virtual environment, the model's control performance under realistic conditions such as hardware latency and signal noise is verified. If the vehicle instability risk reduction rate is ≥90% and there are no system conflicts during the test, the verification requirements are met, and a collaborative pre-control model is finally generated. Specifically, hardware-in-the-loop verification ensures that the model can be transferred from virtual training to actual vehicle control, facilitating subsequent processing.

[0104] Furthermore, the step of determining whether the vehicle has a potential for instability based on the target data using the collaborative pre-control model includes:

[0105] A scene weight matrix is ​​constructed based on the vehicle's driving speed and road surface adhesion coefficient, and the target data is then fused using the scene weight matrix in a scene-based weighted manner to generate a unified data sample.

[0106] The unified data samples are subjected to time-series correlation analysis to extract the instability characteristic parameters of the vehicle.

[0107] The instability characteristic parameters are input into the dynamic decision layer of the collaborative pre-control model to determine whether the vehicle has potential instability based on the instability judgment conditions.

[0108] It's important to note that the first step is to construct a scenario weight matrix. Based on vehicle speed (e.g., low speed <30km / h, medium speed 30-60km / h, high speed >60km / h) and road surface adhesion coefficient (e.g., 0.8 for dry roads, 0.4 for wet roads, and 0.2 for icy roads), matrix elements represent the importance weight of target data for each system in different scenarios. Specifically, for example, in high-speed, low-adhesion scenarios, the wheel speed difference data of the braking system has a higher weight (0.6) than the steering angle data of the steering system (0.3), because high-speed sideslip relies more on braking intervention; in low-speed, high-adhesion scenarios, the steering system data has a higher weight (0.5), because understeer / oversteer is the main risk. This matrix is ​​used to perform scenario-based weighted fusion of target data, generating a unified data sample and eliminating interference from differences in data distribution across different scenarios.

[0109] The second step is time-series correlation analysis and instability feature extraction: The unified data sample is analyzed according to a time series (e.g., continuous data from the most recent 100ms) to extract characteristic parameters reflecting instability trends. Specifically, these include, for example, the "increase in the rate of change of speed difference between the left and right wheels" in the braking system (the faster the increase, the higher the risk of sideslip), the "deviation between the actual steering angle and the ideal steering angle" in the steering system (an increase in deviation increases the risk of understeer / oversteer), and the "difference in the rate of roll angle between the left and right wheels" in the suspension system (the larger the difference, the higher the risk of vehicle rollover). These characteristic parameters not only reflect the current state but also the dynamic trend, providing a basis for early prediction.

[0110] The third step is dynamic decision-making: the instability characteristic parameters are input into the dynamic decision-making layer of the collaborative pre-control model, and the decision-making layer has built-in instability judgment conditions (such as wheel speed difference growth > 50 r / s). 2 The following parameters are considered as potential instability risks: steering angle deviation > 5°, roll rate difference > 2° / s. If any one of these parameters exceeds its corresponding threshold, and the trends of multiple parameters are consistent (e.g., wheel speed difference increase and roll rate difference both exceed the limit simultaneously), the vehicle is judged to have potential instability risks. Specifically, the dynamic decision-making layer uses multi-parameter cross-validation to avoid misjudgment based on a single parameter (e.g., abnormal instantaneous roll rate caused by road bumps). This facilitates subsequent processing.

[0111] Furthermore, the step of inputting the instability characteristic parameters into the dynamic decision layer of the collaborative pre-control model to determine whether the vehicle has potential instability based on the instability judgment conditions includes:

[0112] Based on the differences in the control response parameters of each of the pre-controlled intelligent agents, the instability characteristic parameters are prioritized to generate a corresponding target sequence list;

[0113] According to the target sequence list, each instability feature parameter is sequentially input into the corresponding hierarchical decision node of the dynamic decision layer, and the instability feature parameter received by each hierarchical decision node is subjected to temporal dimension enhancement processing to generate the corresponding temporal enhanced feature sequence.

[0114] The instability trend value corresponding to the vehicle is calculated based on the time-series enhanced feature sequence, and the magnitude of the instability trend value is used to determine whether the vehicle has potential instability risks.

[0115] It should be noted that, firstly, based on the differences in control response parameters of each pre-control agent (such as the braking system responding faster than the suspension system), the instability characteristic parameters are prioritized: for example, in high-speed scenarios, "wheel speed difference increase" (braking system) has the highest priority (because braking intervention is the fastest), followed by "roll angle difference" (suspension system), and finally "steering angle deviation" (steering system). A target sequence list is then generated. Specifically, the prioritization ensures that the decision-making level focuses on the risks that are most easily mitigated through rapid intervention.

[0116] Based on the target sequence list, each instability characteristic parameter is sequentially input into the corresponding hierarchical decision nodes of the dynamic decision layer (e.g., the highest priority parameter is input into the first-level node, and the second highest priority parameter is input into the second-level node). Each hierarchical decision node performs temporal dimension enhancement processing on the received parameters: for example, the instantaneous value of "wheel speed difference increase" is combined with the historical values ​​of the previous 50ms and 100ms to generate a temporally enhanced feature sequence containing time trends (e.g., "increase in speed from 10r / s"). 2 Increase to 50 r / s 2 Time-series enhancement can capture the dynamic changes of parameters and avoid misjudgments based solely on instantaneous values ​​(e.g., a high instantaneous growth rate but a rapid decline may not require intervention).

[0117] Finally, an instability trend value is calculated based on the time-series enhanced feature sequence: using a pre-trained trend prediction model (such as an LSTM neural network), the sequence is mapped to a quantized value of 0-100 (the higher the value, the greater the likelihood of instability). If the trend value exceeds a preset threshold (e.g., 60), an instability risk is identified; if it is below the threshold but shows an upward trend (e.g., from 30 to 55), it is marked as a "potential risk" and continuously monitored. Specifically, this dynamic judgment ensures timely risk detection while avoiding oversensitivity, facilitating subsequent processing.

[0118] Furthermore, the step of outputting multi-system collaborative pre-control instructions adapted to each of the pre-controlled intelligent agents at a preset time includes:

[0119] The vehicle's driving condition data is collected, and a dynamic time algorithm is used to calculate the advance time adapted to the vehicle based on the driving condition data.

[0120] Based on the data acquisition deviation value of each pre-control agent and combined with the training parameters of the collaborative pre-control model, the pre-control command parameter value corresponding to each pre-control agent is determined.

[0121] The multi-system collaborative pre-control command is generated based on the advance time and the pre-control command parameter value.

[0122] It should be noted that the first step is to calculate the advance time: collect vehicle driving condition data (such as vehicle speed, steering angle, and road adhesion coefficient), and estimate the time it takes for the instability trend to develop from "potential danger" to "actual instability" through dynamic time algorithms (such as prediction algorithms based on vehicle dynamics models) (e.g., about 0.5 seconds on a high-speed wet road and about 1 second on a low-speed dry road). This time is the advance time. Specifically, the advance time needs to be long enough to complete the pre-control action, but it cannot be too long to cause unnecessary intervention.

[0123] The second step is to determine the pre-control command parameter values: Each pre-control agent's data acquisition has deviations (e.g., brake pressure sensor error ±2%). Combining these with the training parameters of the collaborative pre-control model (e.g., the model's optimal control parameters under similar working conditions), a compensation algorithm is used to determine the parameter values. Specifically, for example, the pre-control command parameters for the braking agent are "increase the braking force of the left front wheel by 100N ± 5N" (considering sensor deviation), for the steering agent it's "correct the steering angle by + 2° ± 0.1°," and for the suspension agent it's "increase the stiffness of the left side by 50N / mm ± 3N / mm." These parameter values ​​must be within the system's safe range (e.g., braking force does not exceed 80% of the maximum output) to avoid over-control leading to new risks.

[0124] The third step is to generate multi-system collaborative pre-control instructions: Integrate the lead time (e.g., 0.5 seconds) with the parameter values ​​of each agent. The instructions should clearly state "Execute the following actions after 0.5 seconds: Apply 100N of pressure to the left front wheel of the braking system, turn the steering system 2° to the right, and increase the stiffness of the left side of the suspension system by 50N / mm." The instructions must include a timestamp and parameter accuracy range to ensure that all systems execute synchronously and that the magnitude of the actions is controllable, facilitating subsequent processing.

[0125] Furthermore, the step of generating the multi-system collaborative pre-control command based on the advance time and the pre-control command parameter value includes:

[0126] Based on the pre-control execution characteristics of each pre-control agent, the pre-control instruction parameter values ​​are divided into several independent parameter subsets;

[0127] Based on the advance time, and combined with the control response delay of each of the pre-control agents, a corresponding instruction triggering sequence is assigned to each of the independent parameter subsets;

[0128] Each independent parameter subset and its corresponding instruction trigger timing are associated and integrated to generate the multi-system collaborative pre-control instruction.

[0129] It should be noted that, firstly, the parameter values ​​are split according to the pre-control execution characteristics of each pre-control agent: the parameters of the braking system (braking force, pressure), the parameters of the steering system (steering angle, torque), and the parameters of the suspension system (stiffness, damping) constitute independent parameter subsets. Specifically, the splitting ensures that each system only receives parameters related to itself, avoiding command confusion.

[0130] Based on the lead time, and considering the control response delays of each system (e.g., 50ms for the braking system, 30ms for the steering system, and 80ms for the suspension system), the command triggering sequence is allocated. Delay compensation must be considered. For example, the suspension system has the slowest response (80ms), so its command triggering time must be 30ms earlier than the braking system (50ms) to ensure that the actual execution time of all three is consistent (all take effect when the lead time expires). The timing allocation avoids asynchronous interventions caused by differences in system response speeds (e.g., steering has been completed before braking has begun, which may exacerbate instability).

[0131] Finally, the independent parameter subsets are associated and integrated with their corresponding trigger timings: the instruction format includes "system identifier + trigger time + parameter value", for example, "Braking system, T+0.5s, left front wheel braking force 100N; Steering system, T+0.5s, steering angle + 2°; Suspension system, T+0.47s, left side stiffness + 50N / mm" (the suspension system is triggered 30ms in advance). The integrated multi-system coordinated pre-control instructions ensure that each system executes its adaptive actions at the optimal time. Through coordinated intervention of braking, steering, and suspension, instability tendencies are suppressed from multiple dimensions, ultimately maintaining a stable driving state for the vehicle, facilitating subsequent processing.

[0132] Please see Figure 2 The third embodiment of the present invention provides:

[0133] A vehicle stability control system, wherein the system includes:

[0134] The training module is used to configure the braking system, steering system and suspension system inside the vehicle as independent pre-control agents, and to train each of the pre-control agents into a corresponding collaborative pre-control model based on the perception characteristics and control response parameters of each system through a multi-agent learning algorithm.

[0135] The data acquisition module is used to collect corresponding target data through each of the pre-control agents, and to determine whether the vehicle has any potential instability risks based on the target data through the collaborative pre-control model.

[0136] The output module is used to output multi-system collaborative pre-control instructions adapted to each of the pre-control agents at a preset time if the collaborative pre-control model determines that the vehicle has a risk of instability based on the target data.

[0137] The execution module is used to drive the corresponding system to perform pre-control actions through the multi-system collaborative pre-control instructions, so as to keep the vehicle in a stable driving state.

[0138] Furthermore, the training module is specifically used for:

[0139] The braking force output range of the braking system, the steering angle adjustment accuracy of the steering system, and the stiffness response rate of the suspension system are collected respectively, and the corresponding characteristic parameter matrix is ​​established by combining the signal transmission delay of the system sensors.

[0140] The characteristic parameter matrix is ​​fused with vehicle instability sample data under different road conditions, and an independent training network is assigned to each pre-control agent. At the same time, training convergence conditions are set to complete the initial training of each pre-control agent.

[0141] A multi-agent cooperative communication protocol is constructed. Based on the model after initial training, the timing synchronization error and amplitude conflict of the control commands of each pre-control agent are optimized through multiple rounds of iteration to generate the cooperative pre-control model.

[0142] Furthermore, the training module is specifically used for:

[0143] Based on the characteristic parameter matrix, dynamic weight coefficients are constructed, and the control priority of each pre-control agent is allocated under each driving condition of the vehicle through the dynamic weight coefficients.

[0144] During each iteration, the timing of the instruction output of each pre-control agent is adjusted according to the control priority, and the superposition deviation of the control amplitude between different pre-control agents is corrected.

[0145] After each round of iterative optimization, the optimized model is connected to the vehicle hardware-in-the-loop simulation platform and verified based on a preset instability test set until the verification requirements are met, so as to generate the collaborative pre-control model.

[0146] Furthermore, the acquisition module is specifically used for:

[0147] A scene weight matrix is ​​constructed based on the vehicle's driving speed and road surface adhesion coefficient, and the target data is then fused using the scene weight matrix in a scene-based weighted manner to generate a unified data sample.

[0148] The unified data samples are subjected to time-series correlation analysis to extract the instability characteristic parameters of the vehicle.

[0149] The instability characteristic parameters are input into the dynamic decision layer of the collaborative pre-control model to determine whether the vehicle has potential instability based on the instability judgment conditions.

[0150] Furthermore, the acquisition module is specifically used for:

[0151] Based on the differences in the control response parameters of each of the pre-controlled intelligent agents, the instability characteristic parameters are prioritized to generate a corresponding target sequence list;

[0152] According to the target sequence list, each instability feature parameter is sequentially input into the corresponding hierarchical decision node of the dynamic decision layer, and the instability feature parameter received by each hierarchical decision node is subjected to temporal dimension enhancement processing to generate the corresponding temporal enhanced feature sequence.

[0153] The instability trend value corresponding to the vehicle is calculated based on the time-series enhanced feature sequence, and the magnitude of the instability trend value is used to determine whether the vehicle has potential instability risks.

[0154] Furthermore, the output module is specifically used for:

[0155] The vehicle's driving condition data is collected, and a dynamic time algorithm is used to calculate the advance time adapted to the vehicle based on the driving condition data.

[0156] Based on the data acquisition deviation value of each pre-control agent and combined with the training parameters of the collaborative pre-control model, the pre-control command parameter value corresponding to each pre-control agent is determined.

[0157] The multi-system collaborative pre-control command is generated based on the advance time and the pre-control command parameter value.

[0158] Furthermore, the output module is specifically used for:

[0159] Based on the pre-control execution characteristics of each pre-control agent, the pre-control instruction parameter values ​​are divided into several independent parameter subsets;

[0160] Based on the advance time, and combined with the control response delay of each of the pre-control agents, a corresponding instruction triggering sequence is assigned to each of the independent parameter subsets;

[0161] Each independent parameter subset and its corresponding instruction trigger timing are associated and integrated to generate the multi-system collaborative pre-control instruction.

[0162] The fourth embodiment of the present invention provides a computer, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the vehicle stability control method as described above.

[0163] The fifth embodiment of the present invention provides a readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the vehicle stability control method as described above.

[0164] In summary, the vehicle stability control method and system provided by the above embodiments of the present invention can effectively avoid vehicle instability and improve vehicle control efficiency.

[0165] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can reside in the same processor; or the above modules can be located in different processors in any combination.

[0166] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0167] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0168] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0169] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0170] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. A vehicle stability control method, characterized in that, The method includes: The braking system, steering system, and suspension system inside the vehicle are each configured as an independent pre-control agent. Based on the perception characteristics and control response parameters of each system, the pre-control agents are trained into corresponding collaborative pre-control models through a multi-agent learning algorithm. Each of the aforementioned pre-control agents collects corresponding target data, and the collaborative pre-control model determines whether the vehicle has any potential instability risks based on the target data. If the collaborative pre-control model determines that the vehicle has a potential for instability based on the target data, it will output a multi-system collaborative pre-control command adapted to each of the pre-control agents at a preset time. The multi-system collaborative pre-control command drives the corresponding system to perform pre-control actions, so as to keep the vehicle in a stable driving state. The step of training each pre-control agent into a corresponding collaborative pre-control model using a multi-agent learning algorithm based on the perception characteristics and control response parameters of each system includes: The braking force output range of the braking system, the steering angle adjustment accuracy of the steering system, and the stiffness response rate of the suspension system are collected respectively, and the corresponding characteristic parameter matrix is ​​established by combining the signal transmission delay of the system sensors. The characteristic parameter matrix is ​​fused with vehicle instability sample data under different road conditions, and an independent training network is assigned to each pre-control agent. At the same time, training convergence conditions are set to complete the initial training of each pre-control agent. A multi-agent cooperative communication protocol is constructed. Based on the model after initial training, the timing synchronization error and amplitude conflict of the control commands of each pre-control agent are optimized through multiple rounds of iteration to generate the cooperative pre-control model. The step of determining whether the vehicle has a potential for instability based on the target data using the collaborative pre-control model includes: A scene weight matrix is ​​constructed based on the vehicle's driving speed and road surface adhesion coefficient, and the target data is then fused using the scene weight matrix in a scene-based weighted manner to generate a unified data sample. The unified data samples are subjected to time-series correlation analysis to extract the instability characteristic parameters of the vehicle. The instability characteristic parameters are input into the dynamic decision layer of the collaborative pre-control model to determine whether the vehicle has potential instability based on the instability judgment conditions. The step of outputting multi-system collaborative pre-control instructions adapted to each of the pre-control intelligent agents at a pre-preset time includes: The vehicle's driving condition data is collected, and an advance time adapted to the vehicle is calculated based on the driving condition data using a dynamic time algorithm. Based on the data acquisition deviation value of each pre-control agent and combined with the training parameters of the collaborative pre-control model, the pre-control command parameter value corresponding to each pre-control agent is determined. The multi-system collaborative pre-control command is generated based on the advance time and the pre-control command parameter value.

2. The vehicle stability control method according to claim 1, characterized in that, The step of generating the collaborative pre-control model by iteratively optimizing the time synchronization error and amplitude conflict of the control commands of each pre-control agent based on the initialized and trained model includes: Based on the characteristic parameter matrix, dynamic weight coefficients are constructed, and the control priority of each pre-control agent is allocated under each driving condition of the vehicle through the dynamic weight coefficients. During each iteration, the timing of the instruction output of each pre-control agent is adjusted according to the control priority, and the superposition deviation of the control amplitude between different pre-control agents is corrected. After each round of iterative optimization, the optimized model is connected to the vehicle hardware-in-the-loop simulation platform and verified based on a preset instability test set until the verification requirements are met, so as to generate the collaborative pre-control model.

3. The vehicle stability control method according to claim 1, characterized in that, The step of inputting the instability characteristic parameters into the dynamic decision layer of the collaborative pre-control model to determine whether the vehicle has potential instability based on the instability judgment conditions includes: Based on the differences in the control response parameters of each of the pre-controlled intelligent agents, the instability characteristic parameters are prioritized to generate a corresponding target sequence list; According to the target sequence list, each instability feature parameter is sequentially input into the corresponding hierarchical decision node of the dynamic decision layer, and the instability feature parameter received by each hierarchical decision node is subjected to temporal dimension enhancement processing to generate the corresponding temporal enhanced feature sequence. The instability trend value corresponding to the vehicle is calculated based on the time-series enhanced feature sequence, and the magnitude of the instability trend value is used to determine whether the vehicle has potential instability risks.

4. The vehicle stability control method according to claim 1, characterized in that, The step of generating the multi-system collaborative pre-control command based on the advance time and the pre-control command parameter value includes: Based on the pre-control execution characteristics of each pre-control agent, the pre-control instruction parameter values ​​are divided into several independent parameter subsets; Based on the advance time, and combined with the control response delay of each of the pre-control agents, a corresponding instruction triggering sequence is assigned to each of the independent parameter subsets; Each independent parameter subset and its corresponding instruction trigger timing are associated and integrated to generate the multi-system collaborative pre-control instruction.

5. A vehicle stability control system, characterized in that, The system for implementing the vehicle stability control method as described in any one of claims 1 to 4 includes: The training module is used to configure the braking system, steering system and suspension system inside the vehicle as independent pre-control agents, and to train each of the pre-control agents into a corresponding collaborative pre-control model based on the perception characteristics and control response parameters of each system through a multi-agent learning algorithm. The data acquisition module is used to collect corresponding target data through each of the pre-control agents, and to determine whether the vehicle has any potential instability risks based on the target data through the collaborative pre-control model. The output module is used to output multi-system collaborative pre-control instructions adapted to each of the pre-control agents at a preset time if the collaborative pre-control model determines that the vehicle has a risk of instability based on the target data. The execution module is used to drive the corresponding system to perform pre-control actions through the multi-system collaborative pre-control instructions, so as to keep the vehicle in a stable driving state.

6. A computer comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the vehicle stability control method as described in any one of claims 1 to 4.

7. A readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the vehicle stability control method as described in any one of claims 1 to 4.

Images