Vehicle motion control method and device, vehicle and storage medium

Abstract

The present application relates to the technical field of vehicles, and discloses a vehicle motion control method, device, vehicle and storage medium, the method comprising: acquiring motion state information and environment perception information of the vehicle; inputting the motion state information and the environment perception information into a pre-trained neural network model to output reference trajectories at multiple future time points and uncertainty estimates of state quantities of the reference trajectories; acquiring current tire force data from an angle module of the vehicle, and updating at least one model parameter in a vehicle dynamics model based on the tire force data and wheel motion states; using the updated vehicle dynamics model as a prediction model and using the reference trajectories as tracking targets to construct a model predictive control optimization problem; solving the model predictive control optimization problem to obtain optimal control instructions at the current time; and outputting the optimal control instructions to actuators of the vehicle. The method can improve adaptability and robustness under complex working conditions.

Description

Technical Field

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

[0002] With the development of vehicle electrification and intelligence, vehicle motion control systems are facing unprecedented challenges. The corner module configuration introduces redundant actuator degrees of freedom, enabling omnidirectional movements such as crabbing and U-turns, but also significantly increasing control complexity. Current hierarchical control architectures suffer from bottlenecks in model accuracy, rule limitations, and computational burden when handling high-degree-of-freedom, highly nonlinear, and multi-constraint systems. Furthermore, end-to-end neural networks hold great potential for integrated perception and decision-making in autonomous driving, but purely end-to-end methods struggle to guarantee the satisfaction of physical constraints and lack interpretability and safety verification. Summary of the Invention

[0003] This invention aims to at least partially address one of the technical problems in related technologies. To this end, the first objective of this invention is to propose a vehicle motion control method that deeply integrates the intelligent decision-making capabilities of an end-to-end neural network model with the physical constraint handling capabilities of model predictive control. While ensuring physical feasibility, this method enhances the adaptability and robustness of vehicle motion control under complex conditions, retains the neural network model's ability to learn complex driving strategies from massive amounts of data, ensures that the optimal control command always meets the vehicle's physical limits, and automatically adapts to the confidence level of the neural network model's output by dynamically adjusting the tracking weights, achieving efficient solution under redundant degrees of freedom and realizing collaborative optimization.

[0004] The second objective of this invention is to provide a vehicle motion control device.

[0005] The third objective of this invention is to provide a vehicle.

[0006] The fourth objective of this invention is to provide a computer-readable storage medium.

[0007] To achieve the above objectives, a first aspect of the present invention proposes a vehicle motion control method, comprising: acquiring vehicle motion state information and environmental perception information; inputting the motion state information and environmental perception information into a pre-trained neural network model, outputting reference trajectories for multiple future time moments and uncertainty estimates of each state variable of the reference trajectory; acquiring current tire force data from the vehicle's corner module, and updating at least one model parameter in the vehicle dynamics model online based on the tire force data and wheel motion state; constructing a model predictive control optimization problem using the updated vehicle dynamics model as a prediction model and the reference trajectory as a tracking target, wherein the tracking weights of each state variable are dynamically adjusted according to the uncertainty estimates in the cost function of the model predictive control, and the optimization problem also includes vehicle physical constraints; solving the model predictive control optimization problem to obtain the optimal control command at the current time; and outputting the optimal control command to the vehicle's actuator.

[0008] In addition, the vehicle motion control method according to the above embodiments of the present invention may also have the following additional technical features: According to some embodiments of the present invention, obtaining vehicle motion state information includes: fusing collected sensor data through Kalman filtering to estimate the vehicle's longitudinal speed, lateral speed, yaw rate, center of gravity sideslip angle, position coordinates, and heading angle.

[0009] According to some embodiments of the present invention, inputting motion state information and environmental perception information into a pre-trained neural network model includes: organizing motion state information, environmental perception information, and tire force data from multiple past moments into a time series tensor, and using it as input to the neural network model.

[0010] According to some embodiments of the present invention, motion state information and environmental perception information are input into a pre-trained neural network model to output a reference trajectory for multiple future time moments, including: determining the vehicle's desired position, desired heading angle, and desired speed for multiple future time moments based on the motion state information and environmental perception information; and determining the reference trajectory based on the vehicle's desired position, desired heading angle, and desired speed.

[0011] According to some embodiments of the present invention, current tire force data is obtained from the vehicle's corner module, and at least one model parameter in the vehicle dynamics model is updated online based on the tire force data and the wheel motion state, including: acquiring the longitudinal force and lateral force of each wheel from a six-dimensional force sensor installed at the vehicle's corner module, and using the longitudinal force and lateral force as tire force data; acquiring the wheel slip angle and slip ratio, and using the wheel slip angle and slip ratio as the wheel motion state; and updating the stiffness parameter characterizing the lateral mechanical properties of the tire in the vehicle dynamics model according to the tire force data and the wheel motion state.

[0012] According to some embodiments of the present invention, the tracking weight of each state variable is dynamically adjusted based on the uncertainty estimate, including: decreasing the tracking weight of the state variable in the cost function in response to an increase in the uncertainty estimate of the state variable; and increasing the tracking weight of the state variable in the cost function in response to a decrease in the uncertainty estimate of the state variable.

[0013] According to some embodiments of the present invention, solving the model predictive control optimization problem to obtain the optimal control command at the current moment includes: solving for the optimal control sequence at multiple future moments, and taking the first control quantity in the optimal control sequence as the optimal control command at the current moment.

[0014] According to an embodiment of the present invention, a vehicle motion control method includes: acquiring vehicle motion state information and environmental perception information; inputting the motion state information and environmental perception information into a pre-trained neural network model, and outputting reference trajectories for multiple future time moments and uncertainty estimates of each state quantity of the reference trajectory; acquiring current tire force data from the vehicle's corner module, and updating at least one model parameter in the vehicle dynamics model online based on the tire force data and wheel motion state; constructing a model predictive control optimization problem using the updated vehicle dynamics model as a prediction model and the reference trajectory as a tracking target, wherein the tracking weights of each state quantity are dynamically adjusted according to the uncertainty estimates in the cost function of the model predictive control, and the optimization problem also includes vehicle physical constraints; solving the model predictive control optimization problem to obtain the optimal control command at the current time; and outputting the optimal control command to the vehicle's actuator. Therefore, this method can deeply integrate the intelligent decision-making capability of the end-to-end neural network model with the physical constraint processing capability of model predictive control. Under the premise of ensuring physical feasibility, it can improve the adaptability and robustness of vehicle motion control under complex conditions, retain the ability of the neural network model to learn complex driving strategies from massive amounts of data, ensure that the optimal control command always meets the vehicle's physical limits, and automatically adapt to the confidence level of the neural network model output by dynamically adjusting the tracking weights, thereby achieving efficient solution under redundant degrees of freedom and realizing collaborative optimization.

[0015] The second objective of this invention is to propose a vehicle motion control device that deeply integrates the intelligent decision-making capability of an end-to-end neural network model with the physical constraint processing capability of model predictive control. While ensuring physical feasibility, it enhances the adaptability and robustness of vehicle motion control under complex conditions, retains the ability of the neural network model to learn complex driving strategies from massive amounts of data, ensures that the optimal control command always meets the vehicle's physical limits, and automatically adapts to the confidence level of the neural network model output by dynamically adjusting the tracking weights, achieving efficient solution under redundant degrees of freedom and realizing collaborative optimization.

[0016] To achieve the above objectives, a second aspect of the present invention provides a vehicle motion control device, comprising: an acquisition module configured to acquire vehicle motion state information and environmental perception information; a prediction module configured to input the motion state information and environmental perception information into a pre-trained neural network model, and output reference trajectories for multiple future time moments and uncertainty estimates of each state variable of the reference trajectory; an update module configured to acquire current tire force data from the vehicle's corner module, and update at least one model parameter in the vehicle dynamics model online based on the tire force data and wheel motion state; a construction module configured to construct a model predictive control optimization problem using the updated vehicle dynamics model as the prediction model and the reference trajectory as the tracking target, wherein the tracking weights of each state variable are dynamically adjusted according to the uncertainty estimates in the cost function of the model predictive control, and the optimization problem also includes vehicle physical constraints; a solution module configured to solve the model predictive control optimization problem to obtain the optimal control command at the current time; and an execution module configured to output the optimal control command to the vehicle's actuator.

[0017] A vehicle motion control device according to an embodiment of the present invention includes: an acquisition module configured to acquire vehicle motion state information and environmental perception information; a prediction module configured to input the motion state information and environmental perception information into a pre-trained neural network model, and output reference trajectories for multiple future time moments and uncertainty estimates of each state variable of the reference trajectory; an update module configured to acquire current tire force data from the vehicle's corner module, and update at least one model parameter in the vehicle dynamics model online based on the tire force data and wheel motion state; a construction module configured to construct a model predictive control optimization problem using the updated vehicle dynamics model as the prediction model and the reference trajectory as the tracking target, wherein the tracking weights of each state variable are dynamically adjusted according to the uncertainty estimates in the cost function of the model predictive control, and the optimization problem also includes vehicle physical constraints; a solution module configured to solve the model predictive control optimization problem to obtain the optimal control command at the current time; and an execution module configured to output the optimal control command to the vehicle's actuator. Therefore, this device can deeply integrate the intelligent decision-making capability of the end-to-end neural network model with the physical constraint processing capability of model predictive control. Under the premise of ensuring physical feasibility, it can improve the adaptability and robustness of vehicle motion control under complex conditions, retain the ability of the neural network model to learn complex driving strategies from massive amounts of data, ensure that the optimal control command always meets the vehicle's physical limits, and automatically adapt to the confidence level of the neural network model output by dynamically adjusting the tracking weights, thereby achieving efficient solution under redundant degrees of freedom and realizing collaborative optimization.

[0018] To achieve the above objectives, a third aspect of the present invention provides a vehicle comprising: a processor and a memory, wherein the memory stores a program or instructions executable on the processor, and the program or instructions, when executed by the processor, implement the steps of the vehicle motion control method described above.

[0019] According to the embodiments of the present invention, by executing the above-described vehicle motion control method, the vehicle motion control can improve the adaptability and robustness of the vehicle motion control under complex working conditions, retain the ability of the neural network model to learn complex driving strategies from massive amounts of data, ensure that the optimal control command always meets the vehicle's physical limits, automatically adapt to the confidence level of the neural network model output, achieve efficient solution under redundant degrees of freedom, and realize collaborative optimization.

[0020] To achieve the above objectives, a fourth aspect of the present invention provides a computer-readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the vehicle motion control method described above.

[0021] According to the computer-readable storage medium of the present invention, by executing the above-described vehicle motion control method, the adaptability and robustness of vehicle motion control under complex working conditions can be improved, the ability of the neural network model to learn complex driving strategies from massive amounts of data can be preserved, the optimal control command can always meet the vehicle's physical limits, the confidence level of the neural network model output can be automatically adapted, and efficient solution can be achieved under redundant degrees of freedom, thus realizing collaborative optimization.

[0022] 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

[0023] Figure 1 A flowchart of a vehicle motion control method according to some embodiments of the present invention; Figure 2 This is a block diagram of a vehicle motion control device according to some embodiments of the present invention; Figure 3 This is a block diagram of a vehicle according to some embodiments of the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

[0025] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this invention should have the ordinary meaning understood by those skilled in the art. The terms "first," "second," and similar terms used in the embodiments of this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word covers the element or object listed after the word and its equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are only used to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0026] As in the background technology section, with the development of vehicle electrification and intelligence, vehicle motion control systems are facing unprecedented challenges. The corner module configuration (four-wheel independent drive / steering) brings redundant actuator degrees of freedom (4×drive + 4×steering + 4×brake + 4×suspension), enabling vehicles to achieve omnidirectional movements such as crabbing and U-turns, but at the same time, it greatly increases the complexity of control. Current hierarchical control architectures (such as upper-level trajectory planning and lower-level PID (Proportional-Integral-Derivative Controller) / MPC (Model Predictive Control) tracking) face several challenges when dealing with high-degree-of-freedom, highly nonlinear, and multi-constraint systems. These challenges include: model accuracy bottlenecks (lower-level MPC relies on accurate vehicle dynamics models (tire models, suspension models), but these model parameters degrade or even become unstable due to changes in operating conditions (tire wear, road adhesion, load transfer); rule limitations (upper-level trajectory planning is typically based on rules or optimization, making it difficult to handle implicit knowledge in complex dynamic scenarios (such as human driver predictions and game-theoretic behavior)); and computational burden (under redundant degrees of freedom, the optimization problem size of MPC increases dramatically, making real-time solutions difficult). Furthermore, while end-to-end neural networks hold significant potential for integrated perception and decision-making in autonomous driving, purely end-to-end methods struggle to guarantee physical constraint satisfaction and lack interpretability and safety verification.

[0027] In related technologies, vehicle motion control first involves a vehicle state estimation module using Kalman filtering to estimate vehicle speed, center of gravity sideslip angle, and other states. A reference trajectory generation module then generates the desired trajectory based on driver input or an upper-level planner. The MPC controller constructs a finite-time optimization problem based on simplified vehicle dynamics and a tire model to solve for the optimal front wheel steering angle or torque of each wheel. The actuator allocation layer distributes the MPC output to each actuator. The working principle is that in each control cycle, the MPC solves a specific optimization problem based on the current state and the reference trajectory. However, this control method has drawbacks: MPC performance is highly dependent on model accuracy; changes in tire model parameters can easily lead to model mismatch, affecting control performance; redundant drive systems increase the state and control dimensions, drastically increasing the optimization problem-solving time, making it difficult to meet high-speed control requirements; and MPC can only handle explicitly modeled dynamics, making it difficult to accurately describe complex nonlinear and time-varying characteristics.

[0028] Current vehicle motion control employs two main technical approaches: one is an end-to-end method using deep convolutional neural networks, which directly outputs control commands based on camera images or multimodal data as input. Control is achieved through forward propagation after supervised learning training. However, this approach suffers from drawbacks such as lack of interpretability, inability to guarantee physical constraints, extremely high requirements for training data coverage, and difficulty in integration with existing control architectures. The other approach is the MPC method, which excels at guaranteeing physical constraints but relies on an accurate model. Neither approach effectively resolves the contradiction between "intelligent decision-making" and "physical feasibility." Even simple cascading (using end-to-end output commands as MPC references) still suffers from model mismatch and insufficient uncertainty handling.

[0029] Therefore, this invention proposes a deep fusion architecture that utilizes the reference trajectory and uncertainty estimation provided by the neural network model, while simultaneously using measured tire force data to correct the MPC model online, thereby achieving collaborative optimization.

[0030] The vehicle motion control method, vehicle motion control device, vehicle, and computer-readable storage medium proposed in the embodiments of the present invention are described below with reference to the accompanying drawings.

[0031] refer to Figure 1 This is a flowchart of a vehicle motion control method according to some embodiments of the present invention.

[0032] like Figure 1 As shown, the vehicle motion control method of this invention may include the following steps: S101, acquire vehicle motion status information and environmental perception information.

[0033] Specifically, environmental perception information (such as obstacles, lane lines, traffic signs, camera images, LiDAR point clouds, high-precision maps, etc.) of the vehicle can be obtained through environmental perception submodules (such as cameras, LiDAR, millimeter-wave radar, etc.); as well as vehicle motion state information (such as vehicle longitudinal speed, lateral speed, yaw rate, center of gravity sideslip angle, position coordinates, and heading angle, etc.).

[0034] S102 inputs motion state information and environmental perception information into a pre-trained neural network model and outputs reference trajectories for multiple future time points and uncertainty estimates of each state variable of the reference trajectory.

[0035] Specifically, after acquiring the vehicle's motion state information and environmental perception information, the vehicle's motion state information (such as longitudinal velocity, lateral velocity, etc.) and environmental perception information (such as obstacles, lane lines, etc.) from multiple past moments are organized into a specific format and input into a neural network model that has been trained on a large amount of data. This model uses the complex patterns and rules it has learned to analyze and process the input information, and then outputs the vehicle's reference trajectory for multiple future moments. At the same time, it also provides the uncertainty estimate (i.e., the state covariance matrix for each predicted moment) for each state variable (e.g., position, velocity, etc.) on the reference trajectory. , The covariance matrix is ​​a diagonal matrix, with diagonal elements representing the uncertainty variance of each state, thus reflecting the reliability of the prediction results for these state variables. A temporal convolutional network or Transformer can be used, taking the sequence length as input and outputting the trajectory and uncertainty for the next N steps.

[0036] Neural network models can be trained through reinforcement learning and interaction with the environment. The control effect of model predictive control can be used as part of the reward signal, enabling the network to learn to generate reference trajectories that are beneficial for model predictive control tracking.

[0037] In some embodiments, the neural network model can also output the desired control command increment (such as steering wheel angle increment, acceleration increment) and uncertainty estimates of each state quantity of the control command increment. Model predictive control then uses the increment command as a reference while optimizing to ensure that constraints are met. This approach is closer to a direct end-to-end solution but still retains a safety layer for model predictive control.

[0038] S103: Obtain current tire force data from the vehicle's corner module, and update at least one model parameter in the vehicle dynamics model online based on the tire force data and wheel motion state.

[0039] Specifically, a six-dimensional force sensor installed in the vehicle's corner module can collect tire force data such as longitudinal and lateral forces currently acting on each wheel in real time. At the same time, it can acquire motion state information such as wheel slip angle and slip ratio. Using this real-time data, based on specific algorithms and vehicle dynamics principles, at least one key model parameter in the pre-built vehicle dynamics model, which characterizes the lateral mechanical properties of the tire, can be dynamically adjusted and updated to make the model more consistent with the vehicle's current actual operating conditions.

[0040] S104 uses the updated vehicle dynamics model as the prediction model and the reference trajectory as the tracking target to construct a model predictive control optimization problem. In the cost function of model predictive control, the tracking weights of each state variable are dynamically adjusted according to the uncertainty estimate. The optimization problem also includes vehicle physical constraints.

[0041] Specifically, an updated vehicle dynamics model that more accurately reflects the vehicle's current state is used as the prediction model, and the reference trajectories at multiple future moments output by the neural network model are used as the target to be tracked. Based on this, a model predictive control optimization problem is constructed. In this process, considering the uncertainty of each state variable in the reference trajectory, the tracking weights of different state variables are dynamically adjusted in the cost function of model predictive control based on uncertainty estimation. State variables with high uncertainty have lower tracking weights, while those with low uncertainty have higher weights. Simultaneously, to ensure that the control commands conform to the vehicle's actual operating capabilities, physical constraints such as maximum steering angle and maximum acceleration are also incorporated into the optimization problem.

[0042] S105, solve the model predictive control optimization problem to obtain the optimal control command at the current time.

[0043] Specifically, after constructing a model predictive control optimization problem that includes dynamic weight adjustments and vehicle physical constraints, a specific optimization algorithm (such as Pontryagin's minimum principle or the alternating direction multiplier method) is used to solve the problem. Under the premise of satisfying the vehicle physical constraints, the algorithm continuously searches and adjusts the control variables (such as front wheel steering angle and driving torque) until the cost function reaches its minimum. The corresponding combination of control variables at this point represents the optimal control sequence for multiple future time points. The first control variable in the sequence can be selected from it. This serves as the optimal control command applied to the vehicle's actuators at the current moment, enabling precise control of the vehicle.

[0044] This invention can employ one or a combination of the following two acceleration strategies to solve the above-mentioned model predictive control optimization problem: Fast solution based on Pontryagin minimum principle: The original optimization problem is transformed into a two-point boundary value problem using the Pontryagin minimum principle, and then solved quickly using the target method, which is an order of magnitude faster than traditional numerical optimization.

[0045] Explicit Model Predictive Control: For operating conditions with limited parameter variation range, state partitions and corresponding control laws are pre-calculated, and only table lookups are needed online to achieve microsecond-level response, ensuring the real-time performance of the algorithm and meeting the requirements of high-speed control (control cycle ≤ 10ms).

[0046] S106 outputs the optimal control command to the vehicle's actuators.

[0047] Specifically, after solving the model predictive control optimization problem and obtaining the optimal control command at the current moment, a connection can be established with the vehicle's actuator interface through a specific communication interface. The optimal control command, containing precise control parameters (such as steering angle, throttle opening, or braking pressure), is accurately transmitted to the corresponding actuators in each module in the form of electrical or digital signals. This allows the vehicle's actuators to act immediately according to the optimal control command, thereby achieving real-time and precise control of the vehicle's motion state (such as driving direction and speed), ensuring the vehicle travels safely and stably along the expected trajectory. Then, the above steps are repeated.

[0048] Therefore, this invention excels in parameter identification and model adaptability. It can directly use tire force data for parameter identification, eliminating reliance on tire model assumptions. This not only ensures identification accuracy but also enables rapid response. Simultaneously, the constructed model possesses the ability to adaptively adjust to changing operating conditions. Even when road conditions change abruptly, it can quickly adjust, effectively avoiding control failures caused by model mismatch. Deeply integrating the confidence information of the neural network model into the control objective allows for a clear understanding of the reliability of the referenced information, leading to more reasonable trade-offs based on reliability. Furthermore, when the neural network model output is unreliable, it can automatically switch to a conservative mode, comprehensively enhancing safety.

[0049] In some embodiments of the present invention, obtaining vehicle motion state information includes: fusing the collected sensor data through Kalman filtering to estimate the vehicle's longitudinal speed, lateral speed, yaw rate, center of gravity sideslip angle, position coordinates, and heading angle.

[0050] Specifically, to accurately grasp the vehicle's motion status information, various sensors are used to collect relevant data, such as accelerometers, gyroscopes, and wheel speed sensors, to obtain raw information from different dimensions. However, due to the limitations of single sensors, such as noise interference and measurement limitations, directly using this data is not accurate or reliable enough. At this point, the Kalman filter algorithm can be used. It can comprehensively analyze data collected from different sensors, and through two key steps of prediction and updating, it fuses the data to effectively filter out noise and correct errors. This allows for the accurate estimation of key motion status information of the vehicle, such as longitudinal velocity, lateral velocity, yaw rate, sideslip angle, position coordinates, and heading angle, providing a solid data foundation for subsequent vehicle control and decision-making.

[0051] The vehicle's motion state information can be expressed using the following formula:

[0052] in, This indicates the vehicle's motion status information; Indicates lateral velocity; Indicates longitudinal velocity; Indicates yaw rate; Indicates the centroid sideslip angle; Indicates the horizontal position coordinates; Indicates the vertical position coordinates; Indicates the heading angle.

[0053] In some embodiments of the present invention, inputting motion state information and environmental perception information into a pre-trained neural network model includes: organizing motion state information, environmental perception information, and tire force data from multiple past moments into a time series tensor, which is then used as input to the neural network model.

[0054] Specifically, in order to fully utilize historical information during vehicle operation to improve the prediction accuracy of the neural network model, the vehicle's motion state information (such as longitudinal and lateral speeds, reflecting the vehicle's motion characteristics), environmental perception information (such as obstacle positions, lane line shapes, and other surrounding environment-related data), and tire force data (including longitudinal and lateral forces on the tires, reflecting the tire's stress conditions) recorded at multiple past moments are organized in chronological order and constructed into a time series tensor. This time series tensor can comprehensively and systematically present the vehicle's overall state at different times. It is then used as input data to pass to the neural network model that has been trained with a large amount of data in advance, so that the model can perform subsequent analysis and processing.

[0055] In some embodiments of the present invention, motion state information and environmental perception information are input into a pre-trained neural network model to output a reference trajectory for multiple future moments, including: determining the vehicle's desired position, desired heading angle, and desired speed for multiple future moments based on the motion state information and environmental perception information; and determining the reference trajectory based on the vehicle's desired position, desired heading angle, and desired speed.

[0056] Specifically, after inputting the vehicle's current motion state information and environmental perception information into a pre-trained neural network model, the neural network model will perform in-depth analysis and processing of this input information based on the complex rules and patterns it has learned internally. This will predict the ideal position (i.e., desired position), the driving direction (i.e., desired heading angle), and the appropriate driving speed (i.e., desired speed) that the vehicle should be in at multiple future moments. Subsequently, based on these predicted desired positions, desired heading angles, and desired speeds for multiple future moments, a trajectory that can smoothly connect the various desired state points is determined through a specific trajectory generation algorithm or rule. This trajectory is the reference trajectory for the vehicle to follow at multiple future moments.

[0057] The following formula can be used to express the reference trajectory at multiple future moments:

[0058] in, Represents a reference trajectory for multiple future moments; Indicates the desired position of the lateral vehicle; Indicates the desired longitudinal position of the vehicle; Indicates the desired heading angle; Indicates the desired speed; It indicates multiple points in the future.

[0059] In some embodiments of the present invention, current tire force data is obtained from the vehicle's corner module, and at least one model parameter in the vehicle dynamics model is updated online based on the tire force data and wheel motion state. This includes: acquiring the longitudinal force and lateral force of each wheel from a six-dimensional force sensor installed at the vehicle's corner module, and using the longitudinal force and lateral force as tire force data; acquiring the wheel slip angle and slip ratio, and using the wheel slip angle and slip ratio as wheel motion state; and updating the stiffness parameter characterizing the lateral mechanical properties of the tire in the vehicle dynamics model according to the tire force data and wheel motion state.

[0060] Specifically, to make the vehicle dynamics model more closely resemble actual operating conditions, the longitudinal force experienced by each wheel during driving is first collected in real time from the six-dimensional force sensors installed on the corner modules of the vehicle. and lateral force This data is used as tire force data; simultaneously, the wheel slip angle, which reflects the wheel's motion characteristics, is obtained. and slip ratio This serves as information about the wheel's motion state. Next, using this real-time collected tire force data and wheel motion state information, and based on specific algorithms and vehicle dynamics principles, the stiffness parameters in the vehicle dynamics model used to characterize the tire's lateral mechanical properties (such as the front axle equivalent lateral stiffness) are... Rear axle equivalent lateral stiffness or road surface adhesion coefficient Dynamic adjustments and updates are made to enable the vehicle dynamics model to more accurately simulate the vehicle's current dynamic behavior.

[0061] Using the latest collected tire force data, the equivalent lateral stiffness of the front and rear axles is updated by recursive least squares (or by using Kalman filtering, particle filtering, or neural network models to fit the model parameters online).

[0062] Taking lateral stiffness as an example, for the front axle:

[0063]

[0064]

[0065]

[0066] in, This represents the measured lateral force on the two front wheels; express The estimated equivalent lateral stiffness of the front axle at time 1; express The estimated equivalent lateral stiffness of the front axle at time 1; Indicates Kalman gain (scalar); This represents the average measured lateral force of the two front axle wheels; This represents the average front axle side slip angle; Indicates the forgetting factor (e.g., 0.98); express The covariance of the time-time estimation error (scalar); express The covariance of the time-time estimation error (scalar); This represents the identity matrix (in this case, scalar 1).

[0067] The equivalent lateral stiffness of the front axle is estimated in real time using the average value of the two front wheels and a recursive least squares method with a forgetting factor. The updated parameters are immediately used in the prediction model for the next time step, ensuring that the model remains consistent with the current tire / road conditions.

[0068] The vehicle dynamics model can be a hybrid kinematic / dynamic model that includes longitudinal, lateral, and yaw motion, derived from a two-degree-of-freedom single-track model. The vehicle's longitudinal velocity, lateral velocity, yaw rate, sideslip angle, position coordinates, and heading angle are used as state vectors. (i.e., vehicle motion status information), including the total front wheel steering angle of the vehicle. and total longitudinal acceleration As control input (i.e., environmental perception information). The prediction model can be expressed by the following formula:

[0069] in, express The state vector at any given time; express The state vector at any given time; express Time-based control input; This represents the model parameters in the vehicle dynamics model.

[0070] The cost function of model predictive control can be expressed by the following formula:

[0071]

[0072] in, Represents the cost function; This represents the control increment weight matrix; express Time-based control input; The weight matrix can be dynamically adjusted based on the uncertainty estimate of the neural network model output. This represents the control input weight matrix; The variance representing the uncertainty of the lateral position coordinates; The variance representing the uncertainty of the longitudinal position coordinates; The variance representing the uncertainty of the heading angle; The variance representing the uncertainty of velocity; This indicates the adjustment coefficient (e.g., 0.1). Indicates the baseline weight for lateral position tracking; Represents the baseline weight for longitudinal position tracking; Indicates the reference weight for heading angle tracking; Represents the baseline weights for speed tracking; Indicates the baseline weight for lateral velocity tracking; This represents the baseline weight for yaw rate tracking.

[0073] Vehicle physical constraints include: (1) Actuator limiting:

[0074] in, express The total front wheel steering angle of the vehicle at any given moment; This represents the minimum total front wheel steering angle of the vehicle. This represents the maximum total front wheel steering angle of the vehicle; express Total longitudinal acceleration at time t; This represents the minimum value of the total longitudinal acceleration; This represents the maximum value of the total longitudinal acceleration; (2) Tire force constraint (friction ellipse): (Simplified form) in, This represents the average rear axle side slip angle; This indicates the total vertical load (total normal force) on the vehicle. (3) State constraints: such as sideslip angle limit Yaw rate limits, etc. in, This indicates the maximum value of the sideslip angle limit.

[0075] weight matrix With state covariance matrix The relationship is:

[0076] in, The baseline weight matrix; This indicates element-wise multiplication.

[0077] This design possesses excellent adaptive capabilities. When the uncertainty variance of a certain state is large (i.e., the confidence level is low), the weight of that state in the control is reduced accordingly, lowering the controller's tracking requirements for that state. This effectively avoids runaway problems caused by forced tracking due to unreliable reference information. Furthermore, this design innovatively introduces uncertainty into the constraint boundaries of model predictive control. For example, key parameters such as safety distance and lateral acceleration limits can be dynamically adjusted based on uncertainty conditions, further enhancing stability and safety.

[0078] In some embodiments of the present invention, the tracking weights of each state variable are dynamically adjusted based on the uncertainty estimate, including: decreasing the tracking weight of the state variable in the cost function in response to an increase in the uncertainty estimate of the state variable; and increasing the tracking weight of the state variable in the cost function in response to a decrease in the uncertainty estimate of the state variable.

[0079] Specifically, in model predictive control, considering the uncertainty in the prediction of different state variables, the tracking weight of each state variable in the cost function is dynamically adjusted based on its uncertainty estimate to make the control more reasonable and effective. When the uncertainty estimate of a state variable increases, it means that the reliability of the prediction result of that state variable decreases. At this time, its tracking weight in the cost function is reduced, and the strict requirement for accurate tracking of that state variable is lowered. Conversely, if the uncertainty estimate of a state variable decreases, it indicates that its prediction result is more reliable. Then, its tracking weight in the cost function is increased to strengthen the tracking control of that state variable, thereby improving the overall control effect.

[0080] In some embodiments of the present invention, solving the model predictive control optimization problem to obtain the optimal control command at the current moment includes: solving for the optimal control sequence at multiple future moments, and taking the first control quantity in the optimal control sequence as the optimal control command at the current moment.

[0081] Specifically, after constructing the model predictive control optimization problem, an efficient optimization algorithm is used to solve the problem. Under the premise of satisfying the vehicle's physical constraints, the algorithm continuously iterates and adjusts the control variables to search for the optimal control sequence, which minimizes the cost function at multiple future time points. This sequence contains the specific control actions the vehicle should take at different future time points. Considering that actual control can only act on the current time, the first control variable is selected from this optimal control sequence and used as the optimal control command applied to the vehicle's actuators at the current time. This achieves real-time and precise control of the vehicle, ensuring stable driving according to the expected goals.

[0082] In some embodiments, the neural network model can output parameters of a probability distribution (such as a Gaussian mixture model), and the model predictive control is handled using a stochastic model predictive control framework. For four-wheel independent control, distributed model predictive control can be designed, with each corner module having local model predictive control, achieving global optimality through coordination and reducing computational complexity.

[0083] Therefore, this invention achieves an organic unity of intelligence and safety. It retains the intelligent characteristics of neural network models in learning complex driving strategies from massive amounts of data and generating more reasonable reference trajectories. At the same time, it uses model predictive control to ensure that control commands always conform to the vehicle's physical limits, thus guaranteeing safety. Compared with pure end-to-end methods, it has interpretable safety boundaries and can handle more complex scenarios and uncertainties than pure model predictive control. By using tire force data to correct predictive control model parameters online, it achieves model adaptation, solves the model mismatch problem, and ensures that the controller always matches the current tire / road condition. The predictions of the predictive control model are closer to reality, and it can quickly respond to sudden drops in road adhesion coefficient. The system completes model updates, reduces peak lateral deviation, and improves performance under extreme conditions. It possesses uncertainty perception capabilities, dynamically adjusting tracking weights through uncertainty estimation to automatically adapt to the confidence level of the neural network model output. In real-vehicle testing, it remains stable even when the neural network model is disturbed, enhancing robustness. With the adoption of a fast solution algorithm, it exhibits good real-time performance, with the model predictive control solution time controlled within 5ms, meeting the real-time requirements of mass production. It adopts a modular design, allowing the neural network model and model predictive control to be developed and tested independently, facilitating integration with current autonomous driving systems. The neural network model can be iteratively upgraded, and the underlying model predictive control ensures basic safety, facilitating engineering implementation.

[0084] In summary, the vehicle motion control method according to embodiments of the present invention includes: acquiring vehicle motion state information and environmental perception information; inputting the motion state information and environmental perception information into a pre-trained neural network model, outputting reference trajectories for multiple future time moments and uncertainty estimates of each state quantity of the reference trajectory; acquiring current tire force data from the vehicle's corner module, and updating at least one model parameter in the vehicle dynamics model online based on the tire force data and wheel motion state; constructing a model predictive control optimization problem using the updated vehicle dynamics model as a prediction model and the reference trajectory as a tracking target, wherein the tracking weights of each state quantity are dynamically adjusted according to the uncertainty estimates in the cost function of the model predictive control, and the optimization problem also includes vehicle physical constraints; solving the model predictive control optimization problem to obtain the optimal control command at the current time; and outputting the optimal control command to the vehicle's actuator. Therefore, this method can deeply integrate the intelligent decision-making capability of the end-to-end neural network model with the physical constraint processing capability of model predictive control. Under the premise of ensuring physical feasibility, it can improve the adaptability and robustness of vehicle motion control under complex conditions, retain the ability of the neural network model to learn complex driving strategies from massive amounts of data, ensure that the optimal control command always meets the vehicle's physical limits, and automatically adapt to the confidence level of the neural network model output by dynamically adjusting the tracking weights, thereby achieving efficient solution under redundant degrees of freedom and realizing collaborative optimization.

[0085] It should be noted that the method of this embodiment can be executed by a single device, such as a computer or server. The method of this embodiment can also be applied to a distributed scenario, where multiple devices cooperate to complete the task. In such a distributed scenario, one of these devices may execute only one or more steps of the method of this embodiment, and the multiple devices will interact with each other to complete the above method.

[0086] It should be noted that the above description describes some embodiments of the present invention. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims may be performed in a different order than that shown in the above embodiments and still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0087] Corresponding to the above embodiments, the present invention also proposes a vehicle motion control device.

[0088] like Figure 2 As shown, the vehicle motion control device of this embodiment includes: an acquisition module 210, a prediction module 220, an update module 230, a construction module 240, a solution module 250, and an execution module 260.

[0089] The system comprises the following modules: an acquisition module 210, configured to acquire vehicle motion state information and environmental perception information; a prediction module 220, configured to input motion state information and environmental perception information into a pre-trained neural network model, and output reference trajectories for multiple future time points and uncertainty estimates of each state variable of the reference trajectory; an update module 230, configured to acquire current tire force data from the vehicle's corner module, and update at least one model parameter in the vehicle dynamics model online based on the tire force data and wheel motion state; a construction module 240, configured to construct a model predictive control optimization problem using the updated vehicle dynamics model as the prediction model and the reference trajectory as the tracking target, wherein the tracking weights of each state variable are dynamically adjusted according to the uncertainty estimate in the cost function of model predictive control, and the optimization problem also includes vehicle physical constraints; a solution module 250, configured to solve the model predictive control optimization problem to obtain the optimal control command at the current time; and an execution module 260, configured to output the optimal control command to the vehicle's actuator.

[0090] In some embodiments of the present invention, the acquisition module 210 acquires the motion state information of the vehicle, specifically for: fusing the collected sensor data through Kalman filtering to estimate the vehicle's longitudinal speed, lateral speed, yaw rate, center of gravity sideslip angle, position coordinates and heading angle.

[0091] In some embodiments of the present invention, the prediction module 220 inputs motion state information and environmental perception information into a pre-trained neural network model, specifically for: organizing motion state information, environmental perception information and tire force data from multiple past moments into a time series tensor, which is then used as input to the neural network model.

[0092] In some embodiments of the present invention, the prediction module 220 inputs motion state information and environmental perception information into a pre-trained neural network model and outputs a reference trajectory for multiple future moments. Specifically, it is used to: determine the vehicle's expected position, expected heading angle, and expected speed for multiple future moments based on the motion state information and environmental perception information; and determine the reference trajectory based on the vehicle's expected position, expected heading angle, and expected speed.

[0093] In some embodiments of the present invention, the update module 230 obtains the current tire force data from the vehicle's corner module, and updates at least one model parameter in the vehicle dynamics model online based on the tire force data and the wheel motion state. Specifically, it is used to: collect the longitudinal force and lateral force of each wheel from the six-dimensional force sensor installed at the vehicle's corner module, and use the longitudinal force and lateral force as tire force data; obtain the wheel slip angle and slip ratio, and use the wheel slip angle and slip ratio as the wheel motion state; and update the stiffness parameter characterizing the lateral mechanical properties of the tire in the vehicle dynamics model according to the tire force data and the wheel motion state.

[0094] In some embodiments of the present invention, the construction module 240 dynamically adjusts the tracking weight of each state variable according to the uncertainty estimate, specifically for: reducing the tracking weight of the state variable in the cost function in response to an increase in the uncertainty estimate of the state variable; and increasing the tracking weight of the state variable in the cost function in response to a decrease in the uncertainty estimate of the state variable.

[0095] In some embodiments of the present invention, the solution module 250 solves the model predictive control optimization problem to obtain the optimal control command at the current time. Specifically, it is used to: solve for the optimal control sequence at multiple future times, and take the first control quantity in the optimal control sequence as the optimal control command at the current time.

[0096] It should be noted that for details not disclosed in the vehicle motion control device of this embodiment, please refer to the details disclosed in the vehicle motion control method of this embodiment, which will not be repeated here.

[0097] In summary, the vehicle motion control device according to an embodiment of the present invention includes: an acquisition module configured to acquire vehicle motion state information and environmental perception information; a prediction module configured to input the motion state information and environmental perception information into a pre-trained neural network model, and output reference trajectories for multiple future time moments and uncertainty estimates of each state quantity of the reference trajectory; an update module configured to acquire current tire force data from the vehicle's corner module, and update at least one model parameter in the vehicle dynamics model online based on the tire force data and wheel motion state; a construction module configured to construct a model predictive control optimization problem using the updated vehicle dynamics model as the prediction model and the reference trajectory as the tracking target, wherein the tracking weights of each state quantity are dynamically adjusted according to the uncertainty estimates in the cost function of the model predictive control, and the optimization problem also includes vehicle physical constraints; a solution module configured to solve the model predictive control optimization problem to obtain the optimal control command at the current time; and an execution module configured to output the optimal control command to the vehicle's actuator. Therefore, this device can deeply integrate the intelligent decision-making capability of the end-to-end neural network model with the physical constraint processing capability of model predictive control. Under the premise of ensuring physical feasibility, it can improve the adaptability and robustness of vehicle motion control under complex conditions, retain the ability of the neural network model to learn complex driving strategies from massive amounts of data, ensure that the optimal control command always meets the vehicle's physical limits, and automatically adapt to the confidence level of the neural network model output by dynamically adjusting the tracking weights, thereby achieving efficient solution under redundant degrees of freedom and realizing collaborative optimization.

[0098] For ease of description, the above apparatus is described in terms of its functions, divided into various modules. Of course, in implementing this invention, the functions of each module can be implemented in one or more software and / or hardware components.

[0099] The apparatus of the above embodiments is used to implement the corresponding method in any of the foregoing embodiments and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0100] Corresponding to the above embodiments, the present invention also proposes a vehicle.

[0101] refer to Figure 3 The diagram below is a block diagram of a vehicle according to some embodiments of the present invention. It also illustrates a more specific vehicle hardware structure provided in this embodiment. The device may include: a processor 310, a memory 320, an input / output interface 330, a communication interface 340, and a bus 350. The processor 310, memory 320, input / output interface 330, and communication interface 340 are interconnected internally via the bus 350.

[0102] The processor 310 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this specification.

[0103] The memory 320 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 320 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented by software or firmware, the relevant program code is stored in the memory 320 and is called and executed by the processor 310.

[0104] Input / output interface 330 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touch screens, microphones, various sensors, etc., and output devices may include displays, speakers, vibrators, indicator lights, etc.

[0105] The communication interface 340 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0106] Bus 350 includes a pathway for transmitting information between various components of the device, such as processor 310, memory 320, input / output interface 330, and communication interface 340.

[0107] It should be noted that although the above-described device only shows the processor 310, memory 320, input / output interface 330, communication interface 340, and bus 350, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.

[0108] The vehicles described in the above embodiments are used to implement the corresponding methods in any of the foregoing embodiments and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0109] Based on the same inventive concept, corresponding to the methods of any of the above embodiments, the present invention also provides a computer-readable storage medium storing computer instructions for causing a computer to perform the methods of any of the above embodiments.

[0110] The aforementioned computer-readable storage medium can be any available medium or data storage device that a computer can access, including but not limited to magnetic storage (e.g., floppy disks, hard disks, magnetic tapes, magneto-optical disks (MOs), etc.), optical storage (e.g., CDs, DVDs, BDs, HVDs, etc.), and semiconductor storage (e.g., ROMs, EPROMs, EEPROMs, non-volatile memory (NAND flash), solid-state drives (SSDs)).

[0111] The computer instructions stored in the storage medium of the above embodiments are used to cause the computer to perform the methods of any of the above exemplary method sections, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0112] Furthermore, although the operations of the method of the present invention are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all of the operations shown must be performed to achieve the desired result. Rather, the steps depicted in the flowchart may be performed in a different order. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0113] 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.

[0114] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this invention should have the ordinary meaning understood by those skilled in the art. The terms "first," "second," and similar terms used in the embodiments of this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word covers the element or object listed after the word and its equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are only used to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0115] While the spirit and principles of the invention have been described with reference to several specific embodiments, it should be understood that the invention is not limited to the disclosed specific embodiments, and the division of aspects does not imply that features in these aspects cannot be combined for benefit; such division is merely for ease of description. The invention is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims. The scope of the appended claims is to be interpreted in the broadest sense, thereby encompassing all such modifications and equivalent structures and functions.

Claims

1. A vehicle motion control method, characterized in that, include: Acquire vehicle motion status information and environmental perception information; The motion state information and the environmental perception information are input into a pre-trained neural network model, which outputs reference trajectories for multiple future time points and uncertainty estimates of each state quantity of the reference trajectory. The vehicle obtains current tire force data from the vehicle's corner module and updates at least one model parameter in the vehicle dynamics model online based on the tire force data and wheel motion state. Using the updated vehicle dynamics model as the prediction model and the reference trajectory as the tracking target, a model predictive control optimization problem is constructed. In the cost function of model predictive control, the tracking weights of each state variable are dynamically adjusted according to the uncertainty estimate. The optimization problem also includes vehicle physical constraints. Solve the model predictive control optimization problem to obtain the optimal control command at the current time. The optimal control command is output to the vehicle's actuators.

2. The vehicle motion control method according to claim 1, characterized in that, The acquisition of vehicle motion state information includes: By fusing the collected sensor data through Kalman filtering, the vehicle's longitudinal speed, lateral speed, yaw rate, sideslip angle, position coordinates, and heading angle can be estimated.

3. The vehicle motion control method according to claim 1, characterized in that, The step of inputting the motion state information and the environmental perception information into a pre-trained neural network model includes: The motion state information, environmental perception information, and tire force data from multiple past moments are organized into a time series tensor, which is then used as the input to the neural network model.

4. The vehicle motion control method according to claim 3, characterized in that, The step of inputting the motion state information and the environmental perception information into a pre-trained neural network model and outputting reference trajectories for multiple future time points includes: Based on the motion state information and the environmental perception information, determine the vehicle's expected position, expected heading angle, and expected speed at multiple future moments; The reference trajectory is determined based on the vehicle's desired position, the desired heading angle, and the desired speed.

5. The vehicle motion control method according to claim 1, characterized in that, The process of acquiring current tire force data from the vehicle's corner module and updating at least one model parameter in the vehicle dynamics model online based on the tire force data and wheel motion state includes: The longitudinal and lateral forces of each wheel are collected from the six-dimensional force sensors installed at the corner modules of the vehicle, and the longitudinal and lateral forces are used as the tire force data. Obtain the wheel slip angle and slip ratio, and use the wheel slip angle and slip ratio as the wheel motion state; The stiffness parameters characterizing the lateral mechanical properties of the tires in the vehicle dynamics model are updated based on the tire force data and the wheel motion state.

6. The vehicle motion control method according to claim 1, characterized in that, The step of dynamically adjusting the tracking weights of each state variable based on the uncertainty estimate includes: In response to an increase in the uncertainty estimate of the state quantity, the tracking weight of the state quantity in the cost function is reduced; In response to a decrease in the uncertainty estimate of the state quantity, the tracking weight of the state quantity in the cost function is increased.

7. The vehicle motion control method according to claim 1, characterized in that, Solving the model predictive control optimization problem to obtain the optimal control command at the current moment includes: The optimal control sequence for multiple future time points is obtained by solving the problem, and the first control variable in the optimal control sequence is taken as the optimal control command for the current time point.

8. A vehicle motion control device, characterized in that, include: The acquisition module is configured to acquire vehicle motion state information and environmental perception information; The prediction module is configured to input the motion state information and the environmental perception information into a pre-trained neural network model, and output a reference trajectory for multiple future time points and an uncertainty estimate of each state quantity of the reference trajectory. The update module is configured to obtain current tire force data from the vehicle's corner module and update at least one model parameter in the vehicle dynamics model online based on the tire force data and wheel motion state. The construction module is configured to use the updated vehicle dynamics model as the prediction model and the reference trajectory as the tracking target to construct a model predictive control optimization problem. In the cost function of model predictive control, the tracking weights of each state variable are dynamically adjusted according to the uncertainty estimate. The optimization problem also includes vehicle physical constraints. The solution module is configured to solve the model predictive control optimization problem to obtain the optimal control command at the current moment. The execution module is configured to output the optimal control command to the actuators of the vehicle.

9. A vehicle, characterized in that, include: A processor and a memory, the memory storing a program or instructions executable on the processor, the program or instructions, when executed by the processor, implementing the steps of the vehicle motion control method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the steps of the vehicle motion control method as described in any one of claims 1 to 7.

Images