Vehicle control method, vehicle, and computer storage medium

Abstract

This invention discloses a vehicle control method, a vehicle, and a computer storage medium. The vehicle control method includes: identifying driving style using a first neural network based on driver operation data and vehicle state data to output a driving style probability vector; identifying driving scenarios using a second neural network based on environmental perception data and vehicle state data to output a driving scenario probability vector; fusing the driving style probability vector and the driving scenario probability vector to obtain a fused feature vector; constructing a game model based on the fused feature vector, treating driving style preference and driving scenario adaptability as two opposing forces, and solving for a Nash equilibrium to obtain a multi-actuator collaborative control quantity; and controlling the actions of each actuator in the chassis according to the multi-actuator collaborative control quantity. This method improves the personalization and intelligence of the driving experience while ensuring safety, and is applicable to scenarios such as intelligent cockpits and intelligent driving.

Description

Technical Field

[0001] This invention relates to the field of vehicle technology, and in particular to a vehicle control method, a vehicle, and a computer storage medium. Background Technology

[0002] With the rapid development of automotive intelligence and drive-by-wire chassis, vehicle chassis control is evolving from single-mode presets to adaptive and personalized control. Currently, the coordinated control of multiple chassis actuators such as suspension, drive, braking, and steering has become a key technical means to improve vehicle safety, handling, and driving experience.

[0003] In existing technologies, vehicle control schemes based on driving style recognition have emerged. For example, one scheme establishes a kinematic model by clustering historical data and adjusts control commands according to driving style. However, this scheme is only applicable to the single condition of ramp merging and does not involve adaptive adjustment under complex or extreme conditions, nor does it achieve fusion control of multiple chassis actuators. Another scheme identifies driving scenarios and driving intentions based on environmental and vehicle information, but its braking control does not consider the influence of driving style and lacks personalized adjustment capabilities. Yet another scheme is a short-term driving style recognition and mode switching method based on road conditions, which can switch between sport, comfort, and economy modes. However, it does not inherently combine the influence of driving scenarios on chassis control, and may lead to a mismatch between chassis performance and current conditions under extreme conditions.

[0004] In summary, existing technologies either only identify driving style while ignoring scenario adaptability, or only identify driving scenario while ignoring the driver's personalized preferences, lacking a mechanism to deeply integrate the two. Summary of the Invention

[0005] This invention aims to at least solve one of the technical problems existing in the prior art. To this end, one objective of this invention is to propose a vehicle control method that, by fusing the identification of driving style and driving scenario and introducing a game-theoretic control mechanism, achieves optimal collaborative control of multiple chassis actuators, thereby enhancing the personalization and intelligence of the driving experience while ensuring safety.

[0006] The second objective of this invention is to provide a vehicle.

[0007] The third objective of this invention is to provide a computer storage medium.

[0008] To address the aforementioned problems, a first aspect of the present invention provides a vehicle control method, comprising: acquiring driver operation data, vehicle state data, and environmental perception data; based on the driver operation data and the vehicle state data, performing driving style recognition using a first neural network to output a driving style probability vector; based on the environmental perception data and the vehicle state data, performing driving scenario recognition using a second neural network to output a driving scenario probability vector; fusing the driving style probability vector and the driving scenario probability vector to obtain a fused feature vector; constructing a game model based on the fused feature vector, treating driving style preference and driving scenario adaptability as two sides in a game, and solving for a Nash equilibrium to obtain a multi-actuator cooperative control quantity; and controlling the actions of each actuator in the chassis according to the multi-actuator cooperative control quantity.

[0009] The vehicle control method according to embodiments of the present invention simultaneously identifies driving style and driving scenario, and performs game-theoretic control on the fused feature vector obtained by fusing the two. This solves the defects of the single-selection mode in the prior art and realizes for the first time joint modeling and game-theoretic decision-making of driver's subjective preferences and objective scenario adaptability. In particular, when driving style preferences and driving scenario adaptability requirements conflict, the optimal multi-actuator collaborative control quantity that balances safety and experience is obtained by solving the Nash equilibrium through game theory. This avoids extreme control results caused by unidirectional adaptation and maximizes the satisfaction of the driver's personalized needs while ensuring vehicle stability.

[0010] In some embodiments, the driver operation data includes at least one of accelerator pedal opening change rate, braking frequency, and steering wheel angle entropy value; the vehicle state data includes at least one of vehicle speed, yaw rate, longitudinal acceleration, and lateral acceleration; and the environmental perception data includes at least one of road curvature, road surface adhesion coefficient, traffic density, and weather conditions.

[0011] In some embodiments, feature fusion of the driving style probability vector and the driving scenario probability vector is performed to obtain a fused feature vector, including: concatenating the driving style probability vector and the driving scenario probability vector, inputting them into a fully connected network, and outputting the fused feature vector after nonlinear transformation.

[0012] In some embodiments, the driving style preference and the driving scenario adaptability are considered as two parties in a game. A game model is constructed based on the fused feature vector, and a Nash equilibrium is solved to obtain the multi-actuator collaborative control quantity. This includes: defining a strategy space for the chassis multi-actuator control quantity, wherein the strategy space includes at least one of suspension damping, suspension stiffness, drive torque distribution ratio, braking torque distribution ratio, and accelerator pedal characteristic slope; constructing a driving style benefit function and a driving scenario benefit function based on weight coefficients, wherein the weight coefficients are dynamically adjusted by the fused feature vector; using the sum of the driving style benefit function and the driving scenario benefit function having an improvement rate less than a preset threshold during the iteration process as the Nash equilibrium convergence condition, and using an iterative optimal response algorithm to solve the Nash equilibrium to obtain the multi-actuator collaborative control quantity.

[0013] In some embodiments, the method further includes: when the vehicle is in normal operating condition, controlling the action of each actuator of the chassis according to the multi-actuator cooperative control quantity; when the vehicle meets the safety intervention conditions, controlling the vehicle with an emergency control strategy in a preset safety strategy library; and when the multi-actuator cooperative control quantity fails, controlling the vehicle with an emergency control strategy in the preset safety strategy library or the driver's operating instructions.

[0014] In some embodiments, the method further includes: calculating a stability index of the vehicle based on the vehicle state data, the stability index including at least one of yaw rate deviation, center of gravity sideslip angle, and tire slip ratio; and determining whether the vehicle meets the safety intervention conditions based on the stability index.

[0015] In some embodiments, the first neural network is a long short-term memory network, and the second neural network is a convolutional neural network.

[0016] In some embodiments, the method further includes: displaying the execution intensity percentage and game equilibrium state of each actuator of the chassis in real time in the human-computer interaction interface.

[0017] A second aspect of the present invention provides a vehicle, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the at least one processor executes the computer program to implement the vehicle control method described in the above embodiments.

[0018] According to embodiments of the present invention, the vehicle integrates and identifies driving styles and driving scenarios, and introduces a game-theoretic control mechanism to achieve optimal collaborative control of multiple actuators in the chassis, thereby enhancing the personalization and intelligence of the driving experience while ensuring safety.

[0019] A third aspect of the present invention provides a computer storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the vehicle control method described in the above embodiments.

[0020] 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

[0021] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a flowchart of a vehicle control method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of feature fusion according to an embodiment of the present invention; Figure 3 This is a control diagram illustrating security arbitration and redundancy according to an embodiment of the present invention; Figure 4 This is a schematic diagram of a human-computer interaction interface according to an embodiment of the present invention; Figure 5 This is a flowchart of a vehicle control method according to another embodiment of the present invention; Figure 6 This is a structural block diagram of a vehicle according to an embodiment of the present invention. Detailed Implementation

[0022] The embodiments of the present invention are described in detail below. The embodiments described with reference to the accompanying drawings are exemplary. The embodiments of the present invention are described in detail below.

[0023] To address the aforementioned issues, one objective of this invention is to propose a vehicle control method that integrates and identifies driving styles and driving scenarios, and introduces a game-theoretic control mechanism to achieve optimal collaborative control of multiple chassis actuators, thereby enhancing the personalization and intelligence of the driving experience while ensuring safety.

[0024] The following is for reference. Figure 1 A vehicle control method according to an embodiment of the present invention is described, such as... Figure 1 As shown, the method includes at least the following steps S1-S6.

[0025] Step S1: Acquire driver operation data, vehicle status data, and environmental perception data.

[0026] Among them, driver operation data can reflect the driver's subjective intentions, vehicle status data can reflect the current motion response of the vehicle, and environmental perception data can reflect the objective driving conditions outside the vehicle. Based on this, by comprehensively acquiring the above three types of data in a comprehensive, real-time, and multi-dimensional manner, it is possible to achieve a comprehensive perception from subjective intentions to the objective environment, and provide accurate data input for subsequent driving style recognition, driving scenario recognition, and game control, ensuring the real-time nature and accuracy of control decisions.

[0027] Specifically, the above three types of data are obtained in real time through vehicle sensors such as vehicle speed sensors, accelerator pedal position sensors, brake pedal position sensors, steering wheel angle sensors, cameras, and radar.

[0028] Step S2: Based on driver operation data and vehicle status data, a first neural network is used to identify driving style and output a driving style probability vector.

[0029] Specifically, different drivers have different driving habits and preferences, such as aggressive, conservative, and standard driving styles. These style differences directly affect the driver's expectations for chassis control. For example, aggressive drivers prefer quick response and stronger power output, while conservative drivers focus more on smoothness and safety. Therefore, to accurately identify driving styles, this application inputs driver operation data and vehicle state data into a first neural network to identify driving styles and output a driving style probability vector, thus providing a prerequisite for subsequent personalized chassis control.

[0030] Step S3: Based on environmental perception data and vehicle status data, a second neural network is used to identify driving scenarios and output a driving scenario probability vector.

[0031] Specifically, different driving scenarios, such as urban roads, highways, mountain roads, low-friction surfaces, and congested traffic, place drastically different demands on chassis control. For example, low-friction surfaces require more conservative torque output and more aggressive stability control, highways require higher stability and moderate response, while urban roads prioritize agility and comfort. Therefore, driving style preferences alone cannot meet the safety requirements brought about by changing scenarios. Objective scenario recognition is also the foundation for ensuring chassis control safety. Based on this, to accurately identify driving scenarios, this application inputs environmental perception data and vehicle state data into a second neural network to identify driving scenarios and outputs a driving scenario probability vector, providing quantitative scenario feature input for subsequent fusion recognition.

[0032] The neural network mentioned above is a distributed parallel information processing algorithm mathematical model that imitates the behavioral characteristics of animal neural networks. It has self-learning and adaptive capabilities, can be trained with large-scale labeled samples, has strong generalization ability and environmental adaptability, and can stably output accurate probability vectors under different conditions.

[0033] Step S4: Perform feature fusion between the driving style probability vector and the driving scenario probability vector to obtain a fused feature vector.

[0034] Specifically, existing technologies typically employ a single-choice mode, controlling based solely on driving style or driving scenario. This results in control outcomes that either neglect safety or personalization. To address this issue, it is considered that driving style and driving scenario are essentially two different dimensions of information. Driving style reflects the driver's subjective inclinations, while driving scenario reflects objective environmental constraints. Neither can fully describe the complete decision-making conditions required for vehicle control when used alone. Therefore, this application fuses the two to obtain a fused feature vector. This fused feature vector retains the original information of both driving style and driving scenario, while also including the interactive coupling relationship between the two. This provides a more comprehensive and accurate decision-making basis for subsequent game-theoretic control, solving the problem of the lack of a unified decision-making framework when there is a conflict between subjective preferences and objective needs in existing technologies. Furthermore, by integrating the objective environment and subjective intentions to drive the chassis to adaptively adjust performance, vehicle control becomes more intelligent and personalized.

[0035] Step S5: Taking driving style preference and driving scenario adaptability as the two sides of the game, construct a game model based on the fused feature vector, and solve the Nash equilibrium to obtain the multi-actuator collaborative control quantity.

[0036] Specifically, when driving style preferences and driving scenario adaptability requirements are inconsistent—for example, an aggressive driving style encountering a low-friction surface—controlling solely according to driving style preferences, such as excessively pursuing power response, will sacrifice safety; conversely, controlling solely according to scenario adaptability, such as being overly conservative, will sacrifice driving experience. Therefore, this application introduces game theory as a mathematical tool for resolving multi-objective conflicts. Driving style preferences and driving scenario adaptability are treated as two parties in a game, and a Nash equilibrium is sought through non-cooperative or cooperative game theory to obtain the multi-actuator collaborative control quantity. This approach yields a win-win optimal control strategy. Especially when driving style and driving scenario conflict, game theory control can automatically find the optimal balance between safety and experience, avoiding extreme control results caused by unilateral adaptation. For example, in the "sporty style - low-friction surface" conflict scenario, the game result will not be entirely biased towards aggressiveness or conservatism, but will output a moderate control quantity, satisfying the driver's sporty needs as much as possible while ensuring stability.

[0037] Step S6: Control the actions of each actuator in the chassis according to the multi-actuator collaborative control quantity.

[0038] Specifically, the multi-actuator collaborative control quantities obtained from the game theory are decomposed and mapped according to the physical characteristics of each actuator. Control commands are then sent to each actuator controller via CAN (Controller Area Network) bus or Ethernet to achieve operations such as suspension damping / stiffness adjustment, drive / brake torque distribution, and accelerator pedal characteristic adjustment. Each actuator follows an asynchronous update architecture, allocating different control frequencies based on its physical response characteristics, such as slower suspension response or faster drive response. The game theory unit acts as a central coordinating unit for unified scheduling, ensuring coordinated optimization of longitudinal, lateral, and vertical actuators. In this way, the game theory decision results are translated into actual actuator control commands, achieving coordinated control of multiple chassis actuators in the longitudinal, lateral, and vertical directions. This enables the vehicle to maintain stability and good motion response under complex operating conditions, improving the intelligence level of chassis control.

[0039] The specific control method for each actuator can be PID (Proportional-Integral-Derivative) algorithm, LQR (linear quadratic regulator), etc., without specific restrictions.

[0040] The vehicle control method according to embodiments of the present invention simultaneously identifies driving style and driving scenario, and performs game-theoretic control on the fused feature vector obtained by fusing the two. This solves the defects of the single-selection mode in the prior art and realizes for the first time joint modeling and game-theoretic decision-making of driver's subjective preferences and objective scenario adaptability. In particular, when driving style preferences and driving scenario adaptability requirements conflict, the optimal multi-actuator collaborative control quantity that balances safety and experience is obtained by solving the Nash equilibrium through game theory. This avoids extreme control results caused by unidirectional adaptation and maximizes the satisfaction of the driver's personalized needs while ensuring vehicle stability.

[0041] In some embodiments, driver operation data includes at least one of accelerator pedal opening change rate, braking frequency, and steering wheel angle entropy value; vehicle state data includes at least one of vehicle speed, yaw rate, longitudinal acceleration, and lateral acceleration; and environmental perception data includes at least one of road curvature, road surface adhesion coefficient, traffic density, and weather conditions.

[0042] Specifically, the accelerator pedal opening change rate reflects the intensity of the driver's acceleration intention, and is an important indicator for distinguishing between aggressive and conservative driving styles. The larger the accelerator pedal opening change rate, the more aggressive the driving style. Braking frequency reflects the aggressiveness of the driver's following / deceleration; high-frequency braking usually corresponds to a more cautious driving style. Steering wheel angle entropy reflects the randomness and intensity of the driver's steering operation; the larger the steering wheel angle entropy, the more unstable and aggressive the operation, and vice versa. Vehicle speed and yaw rate are the core characteristics of vehicle motion. Vehicle speed directly affects scene recognition, such as distinguishing between highways and cities, while yaw rate reflects the vehicle's transient response characteristics. Longitudinal acceleration and lateral acceleration reflect the vehicle's dynamic response in the longitudinal and lateral directions, respectively, and are key indicators for evaluating handling stability and comfort. Road curvature is used to identify curve scenarios. Road surface adhesion coefficient is used to identify low-adhesion scenarios such as icy and slippery roads. Traffic density is used to identify congested scenarios. Weather conditions are used to comprehensively judge road conditions and driving risks. As shown above, based on the different contributions of each data item to the recognition of driving style and driving scenario, the above data provides sufficient feature input for subsequent driving style recognition and driving scenario recognition, thereby improving the accuracy and robustness of the recognition.

[0043] In some embodiments, the driving style probability vector and the driving scenario probability vector are fused to obtain a fused feature vector, including: concatenating the driving style probability vector and the driving scenario probability vector, inputting them into a fully connected network, and outputting the fused feature vector after nonlinear transformation.

[0044] Specifically, to fully reflect the nonlinear interaction between driving style and driving scenario, this application concatenates the driving style probability vector S and the driving scenario probability vector C during feature fusion to form a joint feature vector [S, C], which is then input into a fully connected network. The fully connected network learns the nonlinear interaction between driving style and driving scenario through nonlinear transformations such as the ReLU activation function (Rectified Linear Unit) to output a fused feature vector. This fused feature vector not only contains information from both driving style and driving scenario but also captures the correlation patterns and coupling relationships between them through the weight matrix of the fully connected network, thus providing a unified decision-making basis for subsequent game control. Therefore, through the nonlinear transformation of the fully connected network, the interactive coupling relationship between driving style and driving scenario is effectively mined, and the output fused feature vector is more expressive and discriminative than simple concatenation, providing higher-quality decision input for subsequent game control.

[0045] For example, refer to Figure 2As shown, a fully connected network can contain 8 to 16 neurons, uses the ReLU activation function, and outputs a fusion feature vector F_fusion=FC([S,C]).

[0046] In some embodiments, driving style preference and driving scenario adaptability are treated as two parties in a game. A game model is constructed based on fused feature vectors, and a Nash equilibrium is solved to obtain the multi-actuator collaborative control quantity. This includes: defining a strategy space for the chassis multi-actuator control quantity, where the strategy space includes at least one of suspension damping, suspension stiffness, drive torque distribution ratio, braking torque distribution ratio, and accelerator pedal characteristic slope; constructing a driving style reward function and a driving scenario reward function based on weight coefficients, wherein the weight coefficients are dynamically adjusted by the fused feature vectors; using the sum of the driving style reward function and the driving scenario reward function having an improvement rate less than a preset threshold during the iteration process as the Nash equilibrium convergence condition, and using an iterative optimal response algorithm to solve the Nash equilibrium to obtain the multi-actuator collaborative control quantity.

[0047] In the strategy space, suspension damping and stiffness determine the vehicle's vertical dynamics, directly affecting handling stability and comfort; the drive torque distribution ratio determines the drive force distribution between the front and rear axles or left and right wheels, affecting the vehicle's acceleration response and yaw characteristics; the braking torque distribution ratio affects braking stability and deceleration response; and the accelerator pedal characteristic slope determines the vehicle's power response sensitivity when the driver depresses the accelerator pedal, serving as a key parameter connecting the driver's intention and the vehicle's response. Therefore, based on the above considerations, this application uses the aforementioned actuator parameters to constitute a complete strategy space for chassis control, in order to search for the optimal combination within it using a game theory algorithm. It should be noted that the value range of each actuator is determined by the actuator's physical limits.

[0048] For the construction of the payoff functions, the driving style payoff function UA ​​uses responsiveness, handling stability, and power performance as positive payoff terms, and safety loss as a negative payoff term, comprehensively reflecting the satisfaction of style preference with the control quantity. The driving scenario payoff function UB uses safety, stability, and comfort as positive payoff terms, reflecting the satisfaction of scenario adaptability with the control quantity. The weight coefficients are dynamically adjusted by the fusion feature vector F_fusion through a neural network module. For example, when driving style features dominate in F_fusion, the weight of UA increases; when driving scenario features dominate, the weight of UB increases. In this way, the above dynamic weight mechanism ensures that the game result can be flexibly adjusted according to the real-time style-scenario state.

[0049] Among these, the specific values ​​for each of the aforementioned benefits, such as responsiveness, handling, power, safety loss, safety, stability, and comfort, can be obtained by professionals driving the vehicle in a standard style and calibrating based on their driving experience; there are no restrictions on this.

[0050] To solve for Nash equilibrium, an iterative optimal response algorithm can be used. Specifically, a strategy combination is initialized, and then one strategy is fixed while the other strategy is optimized, and this process is repeated iteratively. After each iteration, the total payoff of UA+UB is calculated. When the improvement rate of the total payoff is less than a preset threshold, such as 0.01, Nash equilibrium is considered to have been reached. At this point, unilaterally changing the strategy of either side will not improve its own payoff, and this strategy combination is the optimal solution, thus yielding the multi-actuator collaborative control quantity.

[0051] For example, the game theory model considers driving style preference as player A and driving scenario adaptability as player B. Both players engage in a non-cooperative game within the strategy space of the chassis multi-actuator control variables. The payoff functions of each player include multiple dimensions such as response payoff, handling stability payoff, power payoff, and safety payoff, with the weight coefficients of each dimension dynamically adjusted by the fused feature vector. The Nash equilibrium is solved through an iterative optimal response algorithm, meaning that given the opponent's strategy, neither player can increase their own payoff by unilaterally changing their strategy.

[0052] Specifically, the strategy space p for the control variables of the chassis multi-actuator is defined by the following formula: ]; in, This corresponds to suspension damping. This corresponds to suspension stiffness. This corresponds to the rear wheel steering angle. This corresponds to the drive torque distribution ratio. This corresponds to the braking torque distribution ratio. This corresponds to the slope of the accelerator pedal characteristics.

[0053] The driving style benefit function UA ​​= Wa1 × R_response + Wa2 × R_handling + Wa3 × R_power - Wa4 × D_safety, where R_response represents the response benefit, R_handling represents the handling benefit, R_power represents the power benefit, and D_safety represents the safety loss; and the driving scenario benefit function UB = Wb1 × S_safety + Wb2 × S_stability + Wb3 × S_comfort, where S_safety represents the safe driving benefit, S_stability represents the stability value, and S_comfort represents the comfort value. The weight coefficients in these benefit functions, namely Wa1, Wa2, Wa3, Wa4, Wb1, Wb2, and Wb3, are dynamically adjusted by the fused feature vector F_fusion through a neural network. Then, an iterative optimal response algorithm is used, with the convergence condition being that the improvement rate of the sum of UA and UB is less than a preset threshold, which can be expressed as: , where j represents the iteration number. By iteratively solving for the Nash equilibrium, the optimal multi-actuator cooperative control quantity is obtained.

[0054] It should be noted that when solving for Nash equilibrium, it is not limited to using the rate of increase of the sum of the payoff functions being less than a preset threshold as the Nash equilibrium convergence condition. Other convergence conditions can also be set, and there are no specific restrictions on this.

[0055] For example, in a scenario where there is a conflict between a sporty driving style and a low-friction surface, such as when a vehicle is driving on a wet road with a road adhesion coefficient μ=0.35, the driver exhibits aggressive operating characteristics: a large rate of change in the accelerator pedal and a high steering wheel angle entropy value. The driving style identification result is sporty (confidence 85%), and the driving scenario identification result is low-friction surface (confidence 90%). Based on this, the fusion feature vector obtained is [0.85, 0.15, 0.90, 0.10].

[0056] Then, the system performs driving scenario recognition and driving style game theory. Specifically, the initialization process prioritizes driving style, as shown in the following formula: ; Iteration 1: Driving Scenarios Oriented Approach Choose the optimal response and reduce Increase to 0.6 Up to 2.2; Iteration 2: The driving style-oriented approach has been adjusted to the new strategy, reducing... Adjusted to 1.4 Up to 1.2°; Iteration 3: Both sides continue to adjust until convergence, obtaining the equilibrium strategy, i.e., the multi-actuator cooperative control quantity, as shown in the following expression: ; Finally, based on the control effect of the multi-actuator coordinated control, the vehicle yaw rate deviation decreased from the initial 4.8° / s to 2.3° / s, and the tire slip ratio decreased from 0.22 to 0.12. Thus, while ensuring safety, the acceleration response still reached 78% of the driver's requested torque, achieving a balance between safety and user experience.

[0057] In some embodiments, the method of this application further includes: when the vehicle is in normal operating condition, controlling the action of each actuator of the chassis according to the multi-actuator cooperative control quantity; when the vehicle meets the safety intervention conditions, controlling the vehicle with the emergency control strategy in the preset safety strategy library; when the multi-actuator cooperative control quantity fails, controlling the vehicle with the emergency control strategy in the preset safety strategy library or the driver's operation command.

[0058] Specifically, while game-theoretic control can output a control quantity that balances safety and user experience under normal operating conditions, the multi-actuator coordinated control output may fail to meet safety requirements under extreme conditions such as sudden tire blowouts, a sharp drop in road adhesion coefficients, or system failures. Therefore, a safety arbitration and redundant control mechanism is needed to ensure the vehicle remains safe and controllable under all circumstances. Based on these considerations, this application adopts a three-layer redundant channel architecture, referencing... Figure 3 As shown, the first layer is the main channel, where the multi-actuator collaborative control output from the game controller controls the actions of each actuator in the chassis. The second layer is the hot standby channel, which controls the vehicle based on emergency control strategies in the preset safety strategy library, such as pre-stored tire blowout stability control, rollover prevention control, emergency avoidance assist, and low-adhesion stability control. The third layer is the cold standby channel, which controls the vehicle based on the driver's direct control commands. Therefore, under normal operating conditions, the arbitrator selects the game control command from the main channel, i.e., the multi-actuator collaborative control output, and outputs it to the actuators. When the vehicle meets the safety intervention conditions, the arbitrator switches to the hot standby channel, and the emergency control strategies in the preset safety strategy library take over the vehicle. When the main channel detects a fault, such as sensor failure or communication interruption, causing the multi-actuator collaborative control to fail, the arbitrator can prioritize switching to the hot standby channel. If the hot standby channel also fails, it switches to the cold standby channel, where the driver directly controls the vehicle. Thus, through the aforementioned three-layer redundant channels and safety arbitration mechanism, a seamless switch from normal game control to emergency control and then to manual control is achieved, which greatly improves the safety and reliability of the system and ensures that the vehicle can still operate safely under extreme conditions or system failures.

[0059] In some embodiments, the method of this application further includes: calculating the vehicle's stability index based on vehicle state data, the stability index including at least one of yaw rate deviation, center of gravity sideslip angle and tire slip ratio; and determining whether the vehicle meets the safety intervention conditions based on the stability index.

[0060] Specifically, the yaw rate deviation reflects the difference between the vehicle's actual yaw rate and the driver's desired yaw rate; an excessive deviation indicates a tendency for oversteering or understeering. The center of gravity sideslip angle reflects the degree of deviation between the vehicle's posture and its direction of travel; an excessive value indicates that the vehicle is about to sideslip. The tire slip ratio reflects the relative slippage between the tire and the road surface; an excessive value indicates that the tire adhesion is about to saturate. Based on the characteristics of the above parameters, this application calculates the vehicle's stability indices in real time. When any parameter exceeds its corresponding safety threshold, the vehicle is deemed to meet the safety intervention conditions, and the safety arbitration mechanism is triggered. The safety threshold can be determined based on vehicle dynamics simulation and real-vehicle calibration tests. Thus, by calculating multiple stability indices in real time and comparing them with safety thresholds, a quantitative judgment of safety intervention conditions is achieved, ensuring that the safety arbitration mechanism can intervene in a timely manner at the critical moment when the vehicle is about to become unstable, thus preventing accidents.

[0061] In some embodiments, the first neural network is a long short-term memory network, and the second neural network is a convolutional neural network.

[0062] Among these, Long Short-Term Memory (LSTM) networks, on the one hand, through selective memory mechanisms such as forget gates, input gates, and output gates, can distinguish between long-term driving habits and occasional operational fluctuations. For example, a sudden braking may only be a temporary avoidance maneuver, but a combination of continuous rapid acceleration and high-frequency braking reflects an aggressive tendency. On the other hand, compared with traditional RNNs (Recurrent Neural Networks), LSTM cell states can effectively alleviate the gradient vanishing problem that relies on long-term factors, thereby capturing cross-time step behavioral patterns in driving style. Furthermore, it can compress the temporal information of multiple time steps into a fixed-dimensional hidden state vector as the basis for subsequent classification. Therefore, based on the above considerations, this application adopts a Long Short-Term Memory network as the first neural network, thereby utilizing the temporal modeling capabilities of LSTM to effectively capture the temporal variation patterns of driver operation behavior and extract stable style features from historical driving behavior.

[0063] Specifically, driver operation data and vehicle status data are combined into a time sequence. An LSTM network, through its gating structure, extracts features from the input time-series data, uncovering deep features such as the mean and variance of the accelerator pedal change rate, braking frequency, steering wheel angle entropy, and vehicle speed standard deviation. These features are then passed through a fully connected layer and a Softmax (normalized exponential function) classifier to output a driving style probability vector. Thus, by using an LSTM network to extract deep features from time-series driving data, the driver's personalized driving style can be accurately identified and quantitatively expressed as a probability vector, providing quantitative style feature input for subsequent fusion and recognition.

[0064] For example, the hidden layer of the Long Short-Term Memory network has 128 neurons, and the output layer is normalized by the Softmax function to obtain a three-dimensional driving style probability vector: S=[P_radical, P_conservative, P_normal], where P_radical represents the aggressive style probability, P_conservative represents the conservative style probability, and P_normal represents the standard style probability, and the sum of the three is 1.

[0065] Convolutional Neural Networks (CNNs) excel at extracting spatial structural features from multidimensional data. Environmental perception data often exhibits spatial distribution characteristics, such as variations in road surface adhesion coefficients across different regions and the spatial distribution of road curvature. Therefore, CNNs, through the local receptive field of convolutional kernels and pooling operations, can effectively extract these spatial features and model the coupling relationship between the environment and vehicle state. Based on this, this application uses environmental perception data and vehicle state data, employing the output of a CNN through a fully connected layer and a Softmax classifier to obtain a driving scene probability vector. Thus, by using a CNN network to extract spatial features from environmental perception data and vehicle state data, the current driving scene can be accurately identified and quantitatively expressed in the form of a probability vector, providing quantitative scene feature input for subsequent fusion recognition.

[0066] For example, the probability vector of a driving scenario can be represented as C=[P_urban, P_highway, P_mountain, P_low_friction, P_congestion], where P_urban represents the probability of an urban scenario, P_highway represents the probability of a highway scenario, P_mountain represents the probability of a mountain road scenario, P_low_friction represents the probability of a low-friction scenario, and P_congestion represents the probability of a congestion scenario.

[0067] In summary, this application fully leverages the architectural advantages of each network by matching the most suitable neural network structure to different types of data, achieving high-precision recognition of driving style and driving scenario, and providing high-quality input for subsequent fusion recognition.

[0068] In some embodiments, the method of this application further includes: displaying the execution intensity percentage and game equilibrium state of each actuator of the chassis in real time in the human-computer interaction interface.

[0069] Specifically, after solving for the Nash equilibrium, the game controller normalizes the control quantities of each actuator into percentages of execution intensity, such as 85% for suspension damping and 55% for suspension stiffness. Simultaneously, it displays the game equilibrium state between driving style and driving scenario on the interface; for example, a game ratio of 2:1 indicates a stronger style bias. For example, refer to... Figure 4 As shown, the human-machine interface diagram on the PAD displays dynamic graphics and execution intensity of each actuator, the interaction between driving scenarios and driving styles, and the manual lock button. This information is visualized on the PAD screen using graphical methods such as progress bars, pie charts, and radar graphs, allowing the driver to intuitively understand how the vehicle is currently being controlled and why it is being controlled in this way. Therefore, this visual display of the human-machine interface addresses the driver's distrust of complex control logic, reduces doubts about automatic control, and increases the driver's acceptance and satisfaction with the interactive control system. It also enables the driver to make reasonable predictions and interactions based on the displayed information.

[0070] The following is for reference. Figure 5 The vehicle control method of this invention will be illustrated by example, and the specific steps are as follows.

[0071] Step S10: Collect data, including driver operation data, vehicle status data, and environmental perception data.

[0072] Step S11: Identify the driving style probability vector using an LSTM network.

[0073] Step S12: Identify the probability vector of the driving scene using a CNN network.

[0074] Step S13: Feature fusion is performed using a fully connected network.

[0075] Step S14: Output the fused feature vector.

[0076] Step S15: Construct a game theory model.

[0077] Step S16: Solve for the Nash equilibrium to obtain the multi-actuator cooperative control quantity.

[0078] Step S17: Map the multi-actuator collaborative control quantity to each actuator control instruction.

[0079] Step S18: Calculate the vehicle's stability index.

[0080] Step S19: Determine whether the stability index exceeds the safety threshold or whether the main channel is faulty. If the stability index exceeds the safety threshold or the main channel is faulty, proceed to step S20; if the stability index does not exceed the safety threshold, proceed to step S21.

[0081] Step S20: Activate the preset security policy library and switch to the hot standby channel.

[0082] Step S21: Each executor executes the game instructions.

[0083] In step S22, each actuator executes the safety instructions.

[0084] As described above, this application forms a complete technical closed loop of "fusion recognition → game-theoretic decision-making → security arbitration → redundant execution" through the above steps, covering the entire link from perception to execution. Based on the results of driving style recognition and driving style fusion recognition, game theory is used to achieve multi-objective balance and realize the most undesirable driving mode that conforms to the driver's intention, which significantly improves the personalized driving experience. Moreover, the method of this application will also adjust the chassis performance according to different driving scenarios and different driving styles, maximizing the driver's personalized customization under the premise of safety.

[0085] A second aspect of the present invention provides a vehicle, such as Figure 6 As shown, the vehicle 10 includes at least one processor 1 and a memory 2 communicatively connected to at least one processor 1.

[0086] The memory 2 stores a computer program that can be executed by at least one processor 1. When the at least one processor 1 executes the computer program, it implements the vehicle control method of the above embodiment.

[0087] The vehicle can be any type of passenger car or commercial vehicle, especially electrified and intelligent vehicles equipped with a drive-by-wire chassis, such as drive-by-wire steering, brake-by-wire, drive-by-wire, and active suspension. Memory 2 can be integrated into the vehicle domain controller or chassis domain controller and communicate with each actuator via CAN bus or Ethernet.

[0088] According to embodiments of the present invention, the vehicle integrates and identifies driving styles and driving scenarios, and introduces a game-theoretic control mechanism to achieve optimal collaborative control of multiple actuators in the chassis, thereby enhancing the personalization and intelligence of the driving experience while ensuring safety.

[0089] In some embodiments, the vehicle may include a data acquisition module, a fusion recognition module, a game control module, a security arbitration and redundancy module, and a chassis actuator system.

[0090] The system comprises the following modules: a data acquisition module for collecting driver operation data, vehicle status data, and environmental perception data; a fusion recognition module for outputting fused feature vectors, including a driving style recognition submodule, a driving scenario recognition submodule, and a feature fusion submodule; a game theory control module for constructing a game theory model based on the fused feature vectors, solving for the Nash equilibrium control strategy, and outputting multi-actuator collaborative control quantities; a safety arbitration and redundancy module for independently monitoring vehicle stability, providing three layers of redundant execution channels, and taking over control in dangerous conditions or when the main channel fails; and a chassis actuator system including brake-by-wire, drive-by-wire, steering-by-wire, and active suspension for receiving and executing control commands.

[0091] A third aspect of the present invention provides a computer storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the vehicle control method of the above embodiments.

[0092] In the description of this specification, any process or method described in the flowcharts or otherwise herein may be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order according to the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0093] 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-included 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. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), 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). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since 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 a computer memory.

[0094] 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 of the following techniques known in the art, or a combination thereof: 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.

[0095] Those skilled in the art will understand that all or part of the steps of the methods described in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it includes one or a combination of the steps of the method embodiments.

[0096] Furthermore, the functional units in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0097] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

[0098] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example.

[0099] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A vehicle control method, characterized in that, include: Acquire driver operation data, vehicle status data, and environmental perception data; Based on the driver operation data and the vehicle status data, a first neural network is used to identify the driving style and output a driving style probability vector. Based on the environmental perception data and the vehicle status data, a second neural network is used to identify driving scenarios and output a driving scenario probability vector. The driving style probability vector and the driving scenario probability vector are fused to obtain a fused feature vector; Driving style preference and driving scenario adaptability are treated as the two sides of a game. A game model is constructed based on the fused feature vector, and the Nash equilibrium is solved to obtain the multi-actuator collaborative control quantity. The actions of each actuator in the chassis are controlled according to the multi-actuator collaborative control quantity.

2. The vehicle control method according to claim 1, characterized in that, The driver operation data includes at least one of the following: accelerator pedal opening change rate, braking frequency, and steering wheel angle entropy value; the vehicle status data includes at least one of the following: vehicle speed, yaw rate, longitudinal acceleration, and lateral acceleration; and the environmental perception data includes at least one of the following: road curvature, road surface adhesion coefficient, traffic density, and weather conditions.

3. The vehicle control method according to claim 1, characterized in that, The driving style probability vector and the driving scenario probability vector are fused to obtain a fused feature vector, including: The driving style probability vector and the driving scenario probability vector are concatenated and then input into a fully connected network. After nonlinear transformation, the fused feature vector is output.

4. The vehicle control method according to claim 1, characterized in that, The driving style preference and the driving scenario adaptability are used as two sides in a game. A game model is constructed based on the fused feature vector, and the Nash equilibrium is solved to obtain the multi-actuator cooperative control quantity, including: Define a strategy space for chassis multi-actuator control quantities, wherein the strategy space includes at least one of suspension damping, suspension stiffness, drive torque distribution ratio, braking torque distribution ratio, and accelerator pedal characteristic slope; A driving style benefit function and a driving scenario benefit function are constructed based on weight coefficients, wherein the weight coefficients are dynamically adjusted by the fused feature vector. The sum of the driving style benefit function and the driving scenario benefit function is less than a preset threshold during the iteration process as the Nash equilibrium convergence condition. The Nash equilibrium is solved by the iterative optimal response algorithm to obtain the multi-actuator cooperative control quantity.

5. The vehicle control method according to claim 1, characterized in that, The method further includes: When the vehicle is in normal operating condition, the actions of each actuator on the chassis are controlled according to the multi-actuator collaborative control quantity; When the vehicle meets the conditions for safe intervention, the vehicle is controlled using the emergency control strategy in the preset safety strategy library; When the multi-actuator coordinated control fails, the vehicle is controlled by the emergency control strategy in the preset safety strategy library or by the driver's operating commands.

6. The vehicle control method according to claim 5, characterized in that, Also includes: The vehicle stability index is calculated based on the vehicle state data. The stability index includes at least one of yaw rate deviation, center of gravity sideslip angle, and tire slip ratio. The stability index is used to determine whether the vehicle meets the conditions for safe intervention.

7. The vehicle control method according to claim 1, characterized in that, The first neural network is a long short-term memory network, and the second neural network is a convolutional neural network.

8. The vehicle control method according to claim 1, characterized in that, The method further includes: The execution intensity percentage and game equilibrium state of each actuator in the chassis are displayed in real time in the human-computer interaction interface.

9. A vehicle, characterized in that, include: At least one processor; A memory that is communicatively connected to at least one of the processors; The memory stores a computer program that can be executed by at least one of the processors, and when the at least one processor executes the computer program, it implements the vehicle control method according to any one of claims 1-8.

10. A computer storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the vehicle control method according to any one of claims 1-8.

Images