Vehicle control method, vehicle and medium

Abstract

The invention relates to the field of intelligent driving, and particularly discloses a vehicle control method, a vehicle and a medium, and the method comprises the steps: obtaining the spatio-temporal data of an obstacle in a current control period, and determining an obstacle prediction track based on the spatio-temporal data; modeling transverse displacement of the vehicle at different future time points through a higher-order polynomial based on the obstacle prediction trajectory so as to obtain a plurality of to-be-selected transverse movement trajectories; based on the current driving mode and the dynamically updated cost function, determining an initial transverse movement track of the vehicle in the multiple transverse movement tracks to be selected; in response to the detected sudden obstacle, dynamically correcting the initial transverse movement track to obtain a vehicle transverse movement track; and generating a vehicle control instruction based on the vehicle transverse movement track, and controlling the vehicle based on the vehicle control instruction. By dynamically predicting the track of the obstacle and dynamically correcting the initial transverse track by using the rolling time domain, high-precision prediction and real-time re-planning response of the dynamic obstacle can be realized, and the safety is remarkably enhanced.

Description

Technical Field

[0001] This application relates to the field of intelligent driving, specifically to a vehicle control method, a vehicle, and a medium. Background Technology

[0002] In recent years, with the rise of intelligent driving, more and more vehicles have begun to be equipped with intelligent driving assistance systems. These systems can help drivers avoid risks and improve driving safety to a certain extent.

[0003] However, in practical applications, due to the complexity and uncertainty of the road environment, intelligent driving systems still have problems such as slow response to sudden obstacles and poor user experience in the dynamic obstacle avoidance dimension. Summary of the Invention

[0004] In view of the above problems, this disclosure provides a vehicle control method, vehicle, and medium to overcome or at least partially solve the above problems, the technical solution of which is as follows: A vehicle control method, characterized by comprising: acquiring spatiotemporal data of obstacles within the current control cycle; determining a predicted obstacle trajectory based on the spatiotemporal data; modeling the lateral displacement of the vehicle at different future time points using a high-order polynomial based on the predicted obstacle trajectory to obtain multiple candidate lateral movement trajectories; determining an initial lateral movement trajectory of the vehicle among the multiple candidate lateral movement trajectories based on the current driving mode and a dynamically updated cost function; dynamically correcting the initial lateral movement trajectory using rolling temporal programming in response to the detection of a sudden obstacle to obtain a vehicle lateral movement trajectory; generating a vehicle control command based on the vehicle lateral movement trajectory; and controlling the vehicle based on the vehicle control command to make the vehicle laterally move according to the vehicle lateral movement trajectory.

[0005] This application predicts the trajectory of obstacles, enabling early detection of potential collision risks and providing ample time for vehicle lateral movement trajectory planning. It also models the vehicle's lateral displacement using high-order polynomials, accurately describing the vehicle's possible position at future points in time and generating multiple candidate lateral movement trajectories. Combining the current driving mode and a dynamically updated cost function, the system intelligently selects the optimal initial lateral movement trajectory from these options, ensuring driving safety while maximizing driving efficiency and passenger comfort. When a sudden obstacle is detected, the system immediately and dynamically corrects the initial lateral movement trajectory. This real-time response mechanism rapidly adapts to changes in the road environment, effectively preventing collisions. Finally, precise vehicle control commands are generated based on the corrected lateral movement trajectory, achieving safe and smooth lateral movement. This significantly improves the intelligent driving system's ability to cope with complex road environments and enhances driving safety.

[0006] In one example, the step of acquiring spatiotemporal data of obstacles within the current control cycle and determining the predicted trajectory of obstacles based on the spatiotemporal data specifically includes: acquiring spatiotemporal data of all obstacles within a preset range of the vehicle through detection sensors pre-installed on the vehicle; and inputting the spatiotemporal data of all obstacles into the trajectory prediction model to obtain the predicted trajectory of all obstacles.

[0007] This application utilizes sensors on the vehicle to acquire obstacle data in real time within the current control cycle, and then employs an advanced trajectory prediction model to accurately predict the future trajectory of the obstacles. This prediction not only considers the current state of the obstacles but also incorporates their historical motion data and environmental factors, thereby significantly improving the accuracy and reliability of the prediction. In this way, the system can detect potential collision risks in advance, providing ample time for subsequent trajectory planning.

[0008] In one example, the process of modeling the lateral displacement of the vehicle at different future time points using a high-order polynomial based on the obstacle prediction trajectory to obtain multiple candidate lateral movement trajectories specifically includes: determining the distance between each obstacle and the vehicle based on the obstacle prediction trajectory; removing obstacles with distances exceeding a preset threshold to identify conflicting obstacles among all obstacles; determining the conflicting obstacle prediction trajectory and the vehicle's current driving intention; generating a candidate trajectory set for the vehicle based on the conflicting obstacle prediction trajectory and the current driving intention; and modeling the lateral displacement of the vehicle at different future time points in the candidate trajectory set using a high-order polynomial to obtain multiple candidate lateral movement trajectories.

[0009] The system predicts the trajectory of obstacles to determine the distance between each obstacle and the vehicle. Based on a pre-set distance threshold, obstacles exceeding this threshold are eliminated, thus filtering out potentially threatening obstacles from the entire obstacle pool. Generating a candidate trajectory set for the vehicle solely based on the predicted trajectories of these conflict obstacles significantly reduces the computational load of subsequent trajectory planning, improving the system's real-time responsiveness. Next, the system determines the predicted trajectories of these conflict obstacles and, combined with the vehicle's current driving intention, generates a candidate trajectory set for the vehicle. Finally, high-order polynomials are used to model the lateral displacement of the vehicle at different future time points within this candidate trajectory set, resulting in multiple candidate lateral movement trajectories. This modeling method accurately describes the vehicle's possible positions at future time points, providing a rich set of options for subsequent trajectory selection.

[0010] In one example, the step of modeling the lateral displacement of a vehicle at different future time points in the candidate trajectory set using a high-order polynomial to obtain multiple candidate lateral trajectories specifically includes: using a univariate high-order polynomial with time as the independent variable to represent the lateral displacement of the vehicle; obtaining the vehicle's current lateral position, current lateral velocity, and current lateral acceleration information; determining the vehicle's desired lateral position after the end of the current control cycle; and using the current lateral position, current lateral velocity, current lateral acceleration information, desired lateral position, desired lateral velocity, and desired lateral acceleration as constraints of the univariate high-order polynomial; wherein the desired lateral velocity and the desired lateral acceleration are zero.

[0011] By parameterizing lateral displacement using a fifth-order polynomial, the continuity of acceleration and jerk is ensured, eliminating jitter at its source. Furthermore, by incorporating current vehicle state information and desired future state information as constraints, the generated candidate lateral trajectory ensures that it conforms to both the vehicle's current dynamic characteristics and the desired future parking or driving state, thereby improving the accuracy and reliability of trajectory planning. The trajectory generated by the high-order polynomial exhibits continuous and smooth jerk, significantly enhancing ride quality.

[0012] In one example, the cost dimension of the cost function includes at least one of time efficiency dimension, acceleration smoothness dimension, jerkiness comfort dimension, and target deviation distance dimension; before outputting the vehicle's initial lateral movement trajectory from the multiple candidate lateral movement trajectories based on the predicted collision obstacle trajectory, the current driving mode, and the dynamically updated cost function, the method further includes: determining the vehicle's current driving mode; determining the weight range of each cost dimension weight in the cost function according to the current driving mode; and dynamically optimizing the weights of each cost dimension using a particle swarm optimization algorithm to obtain the dynamically updated cost function.

[0013] By comprehensively considering multiple cost dimensions such as time efficiency, acceleration smoothness, jerkiness comfort, and target deviation distance, the merits of candidate lateral movement trajectories can be fully evaluated. Simultaneously, the weights of each cost dimension are dynamically adjusted according to the current driving mode, and particle swarm optimization is used to optimize the weights online or offline, ensuring that the cost function performs optimally under different driving scenarios, thereby selecting the optimal initial lateral movement trajectory.

[0014] In one example, the step of dynamically correcting the initial lateral trajectory using rolling temporal programming in response to the detection of a sudden obstacle to obtain the vehicle's lateral trajectory specifically includes: acquiring sudden obstacle information, including the speed and position of the sudden obstacle; determining a predicted trajectory of the sudden obstacle based on the sudden obstacle information, and generating a dynamically feasible region based on the predicted trajectory; correcting the initial lateral trajectory based on the dynamically feasible region and a cost function to obtain an intermediate lateral trajectory; determining whether the intermediate lateral trajectory satisfies dynamic feasibility; and if it satisfies the dynamic feasibility, using the intermediate lateral trajectory as the vehicle's lateral trajectory.

[0015] When a sudden obstacle is detected, the system can quickly acquire information about the obstacle and predict its future trajectory. Based on this information, the system generates a dynamically feasible region that considers the possible future positions of the vehicle and the obstacle. Then, the system modifies the initial lateral trajectory using a cost function to obtain an intermediate lateral trajectory that satisfies dynamic feasibility. This real-time correction mechanism ensures that the vehicle can react quickly when encountering a sudden obstacle, avoiding collisions.

[0016] In one example, determining whether the intermediate lateral movement trajectory satisfies dynamic feasibility specifically includes: obtaining the lateral acceleration and lateral movement distance corresponding to each trajectory point in the intermediate lateral movement trajectory; determining the maximum vehicle curvature corresponding to the intermediate lateral movement trajectory based on the lateral acceleration and the lateral movement distance; and determining the maximum vehicle curvature when the maximum vehicle curvature is not higher than a preset curvature.

[0017] By verifying whether the kinetic feasibility is met, and then eliminating intermediate lateral movement trajectories that do not meet the kinetic feasibility, it can be ensured that the final selected vehicle lateral movement trajectory is not only theoretically feasible, but can also be executed safely and stably in actual driving, effectively improving the reliability and safety of the intelligent driving system when dealing with emergencies.

[0018] In one example, generating vehicle control commands based on the vehicle's lateral trajectory specifically includes: determining the desired lateral acceleration of the vehicle at the current time point based on the vehicle's lateral trajectory; obtaining the actual lateral acceleration actually performed by the vehicle at the current time point, as well as the steering system delay time; and generating vehicle control commands based on the desired lateral acceleration, the actual lateral acceleration, and the steering system delay time.

[0019] The system determines the desired lateral acceleration at the current time point based on the vehicle's lateral trajectory and, combined with the actual lateral acceleration and steering system delay, generates precise vehicle control commands. This control method ensures the vehicle travels along the predetermined lateral trajectory while considering the system's actual response characteristics, thus improving control accuracy and stability. Unlike existing technologies that often neglect steering system dynamics, leading to a disconnect between planning and execution, this application significantly reduces tracking errors and enhances the practical feasibility of the planning results by incorporating actuator characteristics into the planning layer.

[0020] A vehicle includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the vehicle control method described in any of the above examples.

[0021] A computer-readable storage medium storing computer-executable instructions configured to perform the vehicle control method described in any of the above examples.

[0022] By means of the above technical solution, this disclosure provides a vehicle control method, a vehicle, and a medium. The vehicle's lateral trajectory is represented by a high-order polynomial, and a dynamically updated cost function is introduced to intelligently select the optimal initial lateral trajectory from multiple candidate trajectories. This cost function comprehensively considers multiple dimensions such as time efficiency, acceleration smoothness, abruptness comfort, and target deviation distance, ensuring that the selected trajectory is both safe and efficient. Simultaneously, the weights of each cost dimension in the cost function are dynamically optimized using a particle swarm optimization algorithm, enabling the system to adaptively adjust its trajectory selection strategy according to different driving modes and road environments.

[0023] Furthermore, when the system detects a sudden obstacle, it immediately initiates rolling temporal planning to dynamically correct the initial lateral movement trajectory. During this process, the system acquires the speed and position information of the obstacle, predicts its future trajectory, and generates a dynamically feasible region. Then, based on this dynamically feasible region and a cost function, the initial lateral movement trajectory is corrected in real time to obtain an intermediate lateral movement trajectory. Finally, by determining whether this intermediate lateral movement trajectory meets the dynamic feasibility requirements, it decides whether to use it as the final vehicle lateral movement trajectory. This real-time response mechanism enables the system to quickly adapt to changes in the road environment, effectively preventing collisions.

[0024] Ultimately, based on the corrected lateral trajectory, the system generates precise vehicle control commands and executes these commands through the vehicle's steering system, achieving safe and smooth lateral movement. Throughout the process, the system considers not only the vehicle's dynamic characteristics but also the driver's intentions and passenger comfort, thus significantly improving the intelligent driving system's ability to cope with complex road environments and enhancing driving safety.

[0025] The above description is merely an overview of the technical solution disclosed herein. In order to better understand the technical means of this disclosure and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this disclosure more apparent and understandable, specific embodiments of this disclosure are described below. Attached Figure Description

[0026] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this disclosure. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart illustrating a vehicle control method according to an embodiment of this disclosure is shown; Figure 2 This illustration shows an overall flowchart of a vehicle control method according to an embodiment of the present disclosure; Figure 3 A schematic diagram of a cost function weight dynamic optimization process is shown in an embodiment of this disclosure; Figure 4 A flowchart illustrating a rolling time-domain planning algorithm according to an embodiment of this disclosure is shown; Figure 5 A schematic diagram of a steering delay compensation process according to an embodiment of this disclosure is shown; Figure 6 A schematic diagram of the structure of a vehicle control device according to an embodiment of the present disclosure is shown; Figure 7 A schematic diagram of the structure of a vehicle according to an embodiment of the present disclosure is shown. Detailed Implementation

[0027] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0028] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0029] In recent years, with the rapid development of intelligent driving technology, more and more cars have begun to be equipped with intelligent driving assistance systems. These systems can greatly assist drivers in identifying potential hazards, thereby effectively avoiding traffic accidents and significantly improving driving safety.

[0030] Nevertheless, in real-world driving scenarios, due to the complexity and unpredictability of road environments, intelligent driving systems may still fail to react quickly and promptly when faced with sudden obstacles. This limits their performance to some extent and poses a potential threat to driving safety.

[0031] Specifically, since most vehicles do not integrate a high-precision obstacle motion prediction module, the algorithms inside their intelligent driving systems (such as cubic polynomials and B-splines) are only applicable to static scenes and cannot respond to dynamic obstacles (such as vehicles that suddenly cut in) in real time.

[0032] Furthermore, existing vehicles ignore steering system delays (e.g., the first-order hysteresis characteristic is not modeled), causing the planned trajectory to exceed the steering motor's response capability, resulting in a large actual tracking error.

[0033] Furthermore, the existing cost function does not explicitly optimize jerkiness (e.g., a cubic polynomial only guarantees continuous acceleration, not continuous jerkiness), resulting in discontinuous jerkiness when the vehicle executes trajectory commands, causing passenger discomfort.

[0034] Therefore, this application provides a vehicle control method, a vehicle, and a medium, such as Figure 1 The diagram shown is a schematic flowchart of a vehicle control method provided in one or more embodiments of this specification. This method can be applied to obstacle avoidance during autonomous driving. The process can be executed by a computing device in the relevant field (e.g., a control module installed in the vehicle or a server deployed in the cloud). Certain input parameters or intermediate results in the process can be manually adjusted to help improve accuracy.

[0035] The analysis method involved in the embodiments of this application can be implemented by a terminal device or a server. The server can be a single device or a system composed of multiple devices, i.e., a distributed server. This application does not impose any special limitations on this. For ease of understanding and description, the following embodiments will be described in detail using an in-vehicle system as an example.

[0036] like Figure 1 and Figure 2 As shown, this application provides a vehicle control method, including: S101: Obtain the spatiotemporal data of obstacles within the current control cycle, and determine the predicted trajectory of obstacles based on the spatiotemporal data.

[0037] First, within the current control cycle, the vehicle's infotainment system acquires spatiotemporal data of obstacles within a preset range. Then, based on this data, it predicts the trajectories of these obstacles over a preset time period. These obstacles include both moving and stationary obstacles. For example, in a highway scenario, this could include vehicles suddenly cutting into the lane, slow-moving trucks ahead, and stationary traffic cones on the roadside.

[0038] There are various ways to acquire spatiotemporal data on obstacles. For example, onboard sensors such as LiDAR, cameras, and millimeter-wave radar can be used to obtain spatiotemporal data on obstacles near the vehicle. These sensors can perceive the environmental information around the vehicle in real time and convert it into digital signals for transmission to the vehicle's infotainment system. For example, LiDAR calculates the distance to obstacles by emitting a laser beam and measuring the time it takes for the reflected light to travel. Combined with scanning angle information, it can obtain the obstacle's position and shape in space. Cameras can capture image information around the vehicle and, through image recognition algorithms, can identify different types of obstacles and estimate their position and motion. Millimeter-wave radar uses electromagnetic waves in the millimeter-wave band to detect targets and has the ability to penetrate fog, smoke, and dust, accurately acquiring distance and speed information of obstacles even in adverse weather conditions. After receiving the data from these sensors, the vehicle's infotainment system performs fusion processing to improve the accuracy and reliability of obstacle detection.

[0039] By comprehensively analyzing data from multiple sensors, the limitations of a single sensor can be overcome, reducing false detections and missed detections. After acquiring obstacle information, the vehicle's infotainment system will also predict the obstacles' trajectories using specific algorithms. For example, for moving obstacles, its trajectory over a future period can be predicted using kinematic or dynamic models based on its current speed, acceleration, and direction of motion. For stationary obstacles, although their positions are relatively fixed, the potential impact of changes in the surrounding environment, such as collisions with other vehicles, must be considered. By acquiring obstacle data and their predicted trajectories, the vehicle's infotainment system can anticipate potential hazards around the vehicle, providing crucial information for subsequent trajectory planning and decision-making.

[0040] In one embodiment, when acquiring spatiotemporal data of obstacles, if the vehicle is not equipped with sensors, it can interact with nearby vehicles to obtain the required spatiotemporal data of obstacles through the sensors of those nearby vehicles. This vehicle-to-vehicle interaction can be achieved through vehicle-to-everything (V2X) technology, such as using Dedicated Short Range Communication (DSRC) or cellular networks for data transmission. Nearby vehicles can share obstacle information detected by their own sensors, including position, speed, and acceleration, with the current vehicle in real time. The current vehicle's infotainment system receives this information and performs fusion processing to obtain a more comprehensive and accurate picture of obstacles and their predicted trajectories. In this way, even if the vehicle itself is not equipped with sensors, it can still achieve effective obstacle detection and trajectory prediction by leveraging the perception capabilities of surrounding vehicles.

[0041] Furthermore, when acquiring spatiotemporal data of obstacles from nearby vehicles, it is also necessary to simultaneously acquire the relative positional relationship between the nearby vehicles and the current vehicle. This allows for the adjustment of the spatial coordinates of the obstacles based on the relative positional relationship between the two vehicles. Simultaneously, irrelevant obstacles can be eliminated based on the relative positional relationship. For example, if the current vehicle is directly in front of a nearby vehicle, it will also be considered an obstacle in the spatiotemporal data collected by the nearby vehicles. This obstacle can be eliminated based on the relative positional relationship to avoid interfering with the current vehicle's trajectory planning. After obtaining accurate spatiotemporal data of obstacles and predicted trajectories, the vehicle's system can proceed to the next stage of trajectory modeling.

[0042] In one embodiment, when determining the obstacle prediction trajectory, the spatiotemporal data of all obstacles within a preset range (e.g., within 20m) of the vehicle are first obtained using detection sensors (such as camera sensors, millimeter-wave radar sensors, or lidar sensors) pre-installed on the vehicle. This spatiotemporal data includes both temporal and spatial data. The spatiotemporal data of all obstacles is then input into the trajectory prediction model to obtain the predicted obstacle trajectory for all obstacles.

[0043] In one embodiment, the predicted trajectory model is a Transformer model. This model, based on a self-attention mechanism, can capture long-range dependencies in the spatiotemporal data of obstacles, thereby more accurately predicting the future trajectory of obstacles. In practical applications, the Transformer model can be trained with a large amount of obstacle motion data to optimize its predictive performance. During training, the model continuously adjusts its parameters to minimize the error between the predicted trajectory and the true trajectory. Once training is complete, the model can be applied to real-world vehicle control scenarios, providing reliable obstacle prediction trajectories for the vehicle's infotainment system.

[0044] S102: Based on the obstacle-predicted trajectory, the lateral displacement of the vehicle at different future time points is modeled using a high-order polynomial to obtain a variety of candidate lateral trajectories.

[0045] After obtaining the predicted obstacle trajectory, the lateral displacement of the vehicle at different future time points can be represented by a high-order polynomial, thus yielding multiple candidate lateral movement trajectories. This high-order polynomial can be either a fifth-order or seventh-order polynomial. The candidate lateral movement trajectory refers to a backup lateral movement trajectory, which includes the time point and the corresponding lateral movement distance.

[0046] In one embodiment, when determining candidate lateral movement trajectories, the distance between each obstacle and the vehicle is first determined based on the obstacle prediction trajectories. Obstacles with distances exceeding a preset threshold are then eliminated to identify conflicting obstacles among all obstacles. Conflicting obstacles refer to obstacles that may obstruct the vehicle's movement. Then, the conflicting obstacle prediction trajectories are determined from all obstacle prediction trajectories, along with the vehicle's current driving intention. The current driving intention can be determined by information such as steering wheel angle or accelerator / brake status, reflecting the driver's desired direction and speed trend. Based on the conflicting obstacle prediction trajectories, the vehicle's current driving intention, and the vehicle's own dynamic characteristics, candidate trajectories are generated. These candidate trajectories refer to the trajectories the vehicle might choose. After obtaining the candidate trajectory set, conflict authentication is performed between the candidate trajectories and the conflicting obstacle trajectories to determine whether a collision will occur. Candidate trajectories that do not collide are selected as candidate lateral movement trajectories.

[0047] During the generation of the candidate trajectory set, factors such as road boundaries and traffic rules are also considered to further constrain the generation range of the candidate lateral movement trajectories. For example, if there is a guardrail or curb on the right side of the road, the generated candidate trajectory will not exceed this boundary. Similarly, if traffic rules require vehicles to yield or stop under certain circumstances, these rules will also be incorporated into the modeling process to ensure that the generated candidate trajectories comply with legal requirements.

[0048] Furthermore, after generating the candidate trajectory set, the lateral displacement of the vehicle at different future time points can be accurately modeled using high-order polynomials (such as fifth- or seventh-order polynomials). This modeling process considers constraints such as the vehicle's maximum acceleration, maximum deceleration, and steering capability to ensure that the generated candidate lateral displacement trajectories not only match the vehicle's actual capabilities but also effectively avoid conflict obstacles. Specifically, a univariate high-order polynomial with time as the independent variable is used to represent the vehicle's lateral displacement.

[0049] Taking a fifth-degree polynomial as an example, in the Frenet coordinate system, a fifth-degree polynomial can be represented as follows:

[0050] in, For time The corresponding lateral movement distance, , , , , , These are preset coefficients.

[0051] In addition, constraints should be set for this polynomial. It is understandable that when... When the value is 0, it indicates the current moment, where the vehicle's lateral position, lateral velocity, and lateral acceleration can be obtained through the vehicle's sensors. After the lateral movement ends, the vehicle will continue moving forward. At this point, the vehicle's lateral position should be the desired lateral position, and the lateral velocity and lateral acceleration should be 0.

[0052] Therefore, the constraints of the polynomial can be obtained: Initial state: ( ), horizontal position is lateral velocity is lateral acceleration is .

[0053] Target state: ( ), horizontal position is lateral velocity is lateral acceleration is .

[0054] This modeling approach yields a variety of candidate lateral movement trajectories, each representing a possible lateral movement pattern the vehicle might take at different future points in time. These candidate trajectories provide a rich selection space for subsequent trajectory selection and optimization.

[0055] Next, the system evaluates and selects these candidate lateral movement trajectories using a dynamically updated cost function. This cost function comprehensively considers multiple dimensions such as time efficiency, acceleration smoothness, jerk comfort, and target deviation distance, calculating a comprehensive cost value for each candidate lateral movement trajectory. By comparing the comprehensive cost values ​​of different trajectories, the system can intelligently select the optimal initial lateral movement trajectory.

[0056] S103: Based on the current driving mode and the dynamically updated cost function, determine the initial lateral movement trajectory of the vehicle from the multiple candidate lateral movement trajectories.

[0057] After obtaining the candidate lateral movement trajectories, the optimal initial lateral movement trajectory can be selected from multiple candidate trajectories based on the current driving mode and cost function. The driving mode can be simply divided into Sport mode, Standard mode, and Aggressive mode. It is understandable that different driving modes correspond to different weights in the cost function. For example, when the driving mode is Aggressive mode, the weight of user comfort decreases, but the weight of time increases.

[0058] In one embodiment, the cost dimension of the cost function may include a time efficiency dimension, an acceleration smoothness dimension, a jerk comfort dimension, and a target deviation distance dimension. The cost function can then be described by the following formula:

[0059] in, The cost function value, , , , For weight coefficients that can be updated, The trajectory execution time (optimization goal: the shorter the time, the better). For lateral acceleration (optimization objective: the smaller the acceleration, the smoother the ride). For agility (optimization goal: the lower the agility, the more comfortable). The deviation between the final lateral position and the target position (optimization objective: the smaller the deviation, the better).

[0060] When a target vehicle is detected to have changed lanes more than the preset number (e.g., 4 times) within one minute, it can be considered that the target vehicle is in aggressive mode, and the speed can be appropriately reduced. (Impetus penalty) and (Time penalty) to allow for faster trajectories.

[0061] When the number of lane changes by the target vehicle within one minute is less than the preset number (e.g., 4 times), it can be considered that the target vehicle is in a stable mode, and the speed can be appropriately increased. (Aggression penalty) Limit aggression to improve ride comfort.

[0062] It should be noted that, in addition to determining the driving mode through historical data, the system can also directly obtain the user's driving mode selection command. For example, the user can actively select Sport mode, Standard mode, or Aggressive mode through the vehicle's human-machine interface. Once the user makes a clear selection, the vehicle's system can determine the driving mode based on the user's choice.

[0063] In one embodiment, since driving modes are often not explicitly defined, even in aggressive mode, the level of aggression can vary depending on the user's driving style. Using the same weights for all dimensions in this case would result in a poor user experience. Therefore, even for the same driving mode, the corresponding weights for each dimension will differ. Thus, when updating the weights of each dimension in the cost function, the weight range of each cost dimension must first be determined based on the current driving mode. Then, particle swarm optimization can be used to dynamically optimize the weights of each cost dimension to obtain a dynamically updated cost function.

[0064] Specifically, the particle swarm optimization algorithm process is as follows: Figure 3 As shown, S301: The particle swarm needs to be initialized. In this step, N particles are randomly generated, and each particle represents a set of weights. ,and That is, the magnitude of the weight corresponding to each particle is all located in Inside.

[0065] S302: Calculate the fitness of each of the N particles, i.e., for each particle... Calculate its cost function value The method for calculating the cost function value has been given above and will not be repeated here.

[0066] S303: After determining the cost function value for each particle, each particle records its historical optimal solution to update its individual optimal solution. , .in, For the first The individual optimal solution after the next iteration. Let A be a function that represents the set of values ​​of the independent variable that minimizes the objective function within the parentheses; This refers to the particle swarms of each generation during the iteration process.

[0067] S304: After updating the individual optimal solution, all particles share the current optimal solution to update the global optimal solution. , .

[0068] S305: Then each particle moves towards its individual optimal and global optimal direction to update its velocity and position, specifically:

[0069]

[0070] in, Inertial weight (controlling the inertia of particle motion). , The learning factor (the attractiveness that controls individual / global optimality). , For random disturbance terms, For the first Particle velocity at the next iteration.

[0071] After updating the particle velocity and position, S306: Determine whether the termination condition is met. The termination condition can be set to reach the preset maximum number of iterations or fitness convergence.

[0072] When the termination condition is met, S307: Output the optimal cost dimension weight corresponding to the particle. If the termination condition is not met, repeat S302 until the termination condition is met.

[0073] S104: In response to the detection of a sudden obstacle, dynamically correct the initial lateral trajectory to obtain the vehicle lateral trajectory.

[0074] In the lateral motion planning of autonomous driving, when a sudden obstacle is detected (such as a vehicle suddenly cutting in), the initial lateral trajectory needs to be locally adjusted to obtain the vehicle's lateral trajectory, so as to ensure real-time safety and feasibility.

[0075] In one embodiment, such as Figure 4 As shown, if a sudden obstacle is detected, Receding Horizontal Planning (RHP) can be used to dynamically correct the initial lateral movement trajectory. Specifically, after detecting a sudden obstacle, information about the obstacle, including its speed and position, is acquired. Then, based on this information, a predicted trajectory is determined, and a dynamically feasible region is generated. Next, based on the dynamically feasible region and the cost function, the initial lateral movement trajectory is corrected to obtain an intermediate lateral movement trajectory. It is then determined whether the intermediate trajectory satisfies dynamic feasibility. If it does, the intermediate trajectory is used as the vehicle's lateral movement trajectory. If it does not satisfy dynamic feasibility, a preset emergency strategy is triggered.

[0076] More specifically, when a sudden obstacle is detected, the environmental state is updated, and the vehicle-mounted system acquires the latest sensor data (i.e., obstacle position, speed, etc.). Based on the obstacle's position and speed, it predicts the obstacle's trajectory and updates the dynamic constraints in the Frenet coordinate system. Then, a dynamically feasible region (safe corridor) is generated based on the predicted obstacle trajectory. This dynamically feasible region is... ,in, The safe corridor function, whose corresponding value is dynamically adjusted based on the predicted obstacle trajectory, is a time-dependent function. A set, this set contains time All safe lateral positions. The lateral position planned for the vehicle's driving trajectory. This is the minimum allowable lateral position for a vehicle (i.e., the rightmost boundary). This represents the maximum allowed lateral position of the vehicle, i.e., the leftmost boundary.

[0077] After generating the safe corridor, within the time window Within this process, the initial lateral trajectory is corrected based on a fifth-order polynomial to obtain an intermediate lateral trajectory, minimizing the constraints on new obstacle avoidance, actuator dynamics, and comfort. The cost function can then be expressed as:

[0078] After generating the intermediate lateral movement trajectory, it is necessary to verify whether the trajectory satisfies dynamic feasibility, i.e., whether the curvature exceeds the vehicle's limits. This can be expressed by the following formula:

[0079] in, for Vehicle curvature at any given moment. This represents the maximum vehicle curvature. If dynamic feasibility is satisfied, the optimal trajectory for the current moment is executed, and the time window is slid to the next control cycle.

[0080] S105: Generate vehicle control commands based on the vehicle lateral movement trajectory, and control the vehicle based on the vehicle control commands so that the vehicle moves laterally according to the vehicle lateral movement trajectory.

[0081] After obtaining the vehicle's lateral movement trajectory, vehicle control commands can be generated based on the trajectory. The vehicle control system can then control the vehicle using these commands, enabling it to move laterally according to the lateral movement trajectory.

[0082] In one embodiment, such as Figure 5 As shown, when generating vehicle control commands, due to actuator delays within the vehicle, the desired lateral acceleration of the vehicle at the current time point can be determined based on the vehicle's lateral trajectory; the actual lateral acceleration executed by the vehicle at the current time point, as well as the steering system delay time, can be obtained; based on the desired lateral acceleration, the actual lateral acceleration, and the steering system delay time, vehicle control commands are generated. Specifically, the steering delay compensation formula can be expressed as:

[0083] in, For the target lateral acceleration, The steering system delay time constant, For the desired lateral acceleration, This refers to the vehicle's actual lateral acceleration. When generating control commands, the vehicle will... Generate and avoid steering delays within the vehicle.

[0084] The aforementioned preset thresholds, such as the preset number of times, can be pre-stored in the computer device's storage. When vehicle control commands need to be generated, the computer device can select the preset thresholds from the storage. Alternatively, the computer device can obtain these preset thresholds from other external devices. For example, the preset thresholds can be stored in the cloud, and the computer device can retrieve them from the cloud when vehicle control commands need to be generated. This embodiment does not limit the method of obtaining the preset thresholds.

[0085] This application introduces the Transformer deep learning model to predict the future trajectories of surrounding dynamic obstacles in real time, replacing the traditional fixed trajectory assumption. Based on the prediction results, a dynamic safety corridor is then constructed in the Frenet coordinate system. A rolling temporal planning framework is adopted. Within each control cycle (e.g., 50ms), based on the latest environmental information, the local trajectory for the next short time domain (e.g., within 3s) is replanned within the safety corridor. This completely solves the problem of slow response to dynamic obstacles in traditional algorithms and can effectively cope with complex scenarios such as sudden vehicle intrusion. The dynamic safety corridor ensures the geometric feasibility of obstacle avoidance; the rolling temporal planning framework can decompose complex problems into a series of short-time domain optimization problems, with low computational cost, meeting the real-time requirements of the vehicle platform.

[0086] Furthermore, in the Frenet system's lateral planning, this application employs a fifth-order polynomial d(t) for parametric modeling of lateral displacement. This model inherently guarantees the continuity of acceleration and jerk force, eliminating jitter at its source. Simultaneously, a comprehensive cost function incorporating time efficiency, acceleration smoothness, jerk force comfort, and terminal error is designed, and the weight coefficients of this cost function are innovatively optimized offline or online using a particle swarm optimization algorithm. The trajectory generated by the fifth-order polynomial exhibits continuous and smooth jerk force, significantly improving ride quality. The particle swarm optimization algorithm automatically optimizes different weight combinations for "aggressive mode" and "comfort mode" (e.g., automatically assigning higher weights to the Jerk term in comfort mode), replacing tedious and subjective manual parameter tuning, improving optimization efficiency by over 80%, and enabling a single algorithm to adapt to various scenario requirements.

[0087] Finally, this application establishes a first-order hysteresis dynamic model of the steering system. Before sending the planned desired lateral acceleration to the controller, this model is used for feedforward compensation to calculate a compensated control command that better matches the actual system response. This effectively overcomes actuator delay, enabling the vehicle to track the planned trajectory more accurately, and reducing the measured control error by 30%. Simultaneously, the vehicle state estimate (actual acceleration) is fed back to the compensator, forming a closed-loop optimization of perception, planning, execution, and feedback, enhancing the system's robustness.

[0088] In summary, in the dynamic obstacle avoidance dimension, compared to existing technologies that are mostly based on static environments or simple dynamic assumptions, the response is lagging. This application achieves high-precision prediction of dynamic obstacles and millisecond-level real-time replanning response through Transformer prediction, dynamic safe corridor planning, and rolling temporal planning, significantly enhancing safety.

[0089] In terms of comfort, compared to the discontinuous cubic polynomial jerkiness of existing technologies, which causes vibration, MPC comfort relies on complex parameter tuning. This application ensures the ultimate smoothness of the trajectory in principle by using a continuous quintic polynomial of jerkiness and adaptive weight optimization of the cost dimension. By using AI optimization to adaptively match different driving styles, both comfort and efficiency are achieved.

[0090] In terms of actuator constraints, compared to existing technologies that often neglect steering system dynamics and disconnect planning from execution, this application incorporates actuator characteristics into the planning layer through feedforward compensation using a first-order lag model, which significantly reduces tracking errors and improves the practical feasibility of the planning results.

[0091] In terms of system integration, compared with the relatively independent modules and weak collaboration of existing technologies, this application adopts a multi-module deep integration (prediction, optimization, planning, compensation) architecture to build an end-to-end, closed-loop horizontal active safety planning system, which has better overall performance than the simple superposition of each part.

[0092] It should be noted that the above algorithms are only illustrative examples and are not intended to limit the algorithms that can achieve the above functions. For example, the particle swarm optimization algorithm can be replaced by a genetic algorithm, which has a slower convergence speed but stronger global optimization ability. The Transformer prediction model can be replaced by a long short-term memory neural network, which has slightly lower accuracy but lighter computation.

[0093] In practical applications, algorithms can be flexibly selected and adjusted according to specific needs and hardware conditions. For example, on resource-constrained vehicle platforms, a combination of algorithms with low computational cost and good real-time performance may be preferred; while in scenarios with extremely high safety requirements, the accuracy and robustness of the algorithm may be more important, and some computational efficiency may be sacrificed for better performance.

[0094] Furthermore, the vehicle control method proposed in this application is not only applicable to passenger vehicles, but can also be extended to various vehicle types and scenarios such as commercial vehicles, autonomous taxis, and logistics vehicles. By adjusting the algorithm parameters and model structure, it can quickly adapt to the dynamic characteristics and usage requirements of different vehicles, demonstrating good versatility and scalability.

[0095] In addition, such as Figure 6 As shown, Figure 6 This is a schematic diagram of a vehicle control device provided in an embodiment of this application. The device includes: The environmental perception module 601 acquires spatiotemporal data of obstacles within the current control cycle and determines the predicted trajectory of the obstacles based on the spatiotemporal data. The trajectory modeling module 602, based on the obstacle-predicted trajectory, models the lateral displacement of the vehicle at different future time points using a high-order polynomial to obtain a variety of candidate lateral trajectories; The trajectory planning module 603 determines the initial lateral movement trajectory of the vehicle from the multiple candidate lateral movement trajectories based on the current driving mode and the dynamically updated cost function. The dynamic correction module 604, in response to the detection of a sudden obstacle, dynamically corrects the initial lateral trajectory using rolling time-domain planning to obtain the vehicle's lateral trajectory; The instruction generation module 605 generates vehicle control instructions based on the vehicle's lateral movement trajectory, and controls the vehicle based on the vehicle control instructions so that the vehicle moves laterally according to the vehicle's lateral movement trajectory.

[0096] In one specific embodiment, the environmental perception module 601 includes: acquiring spatiotemporal data of all obstacles within a preset range of the vehicle through detection sensors preset on the vehicle; and inputting the spatiotemporal data of all obstacles into a trajectory prediction model to obtain the obstacle prediction trajectory of all obstacles.

[0097] In one specific embodiment, the trajectory modeling module 602 includes: determining the distance between each obstacle and the vehicle based on the obstacle prediction trajectory; removing obstacles whose distance is higher than a preset threshold to identify conflict obstacles among all obstacles; determining the conflict obstacle prediction trajectory and the vehicle's current driving intention; generating a candidate trajectory set for the vehicle based on the conflict obstacle prediction trajectory and the current driving intention; and modeling the lateral displacement of the vehicle in the candidate trajectory set at different future time points using a high-order polynomial to obtain multiple candidate lateral movement trajectories.

[0098] In one specific embodiment, the trajectory modeling module 602 includes: representing the lateral displacement of the vehicle using a univariate high-order polynomial with time as the independent variable; acquiring the vehicle's current lateral position, current lateral velocity, and current lateral acceleration information; determining the vehicle's desired lateral position after the end of the current control cycle; and using the current lateral position, current lateral velocity, current lateral acceleration information, desired lateral position, desired lateral velocity, and desired lateral acceleration as constraints on the univariate high-order polynomial; wherein the desired lateral velocity and the desired lateral acceleration are zero.

[0099] In one specific embodiment, the cost dimension of the cost function includes at least one of time efficiency dimension, acceleration smoothness dimension, jerkiness comfort dimension, and target deviation distance dimension; the trajectory planning module 603 includes: determining the current driving mode of the vehicle; determining the weight range of each cost dimension weight in the cost function according to the current driving mode; and dynamically optimizing the weights of each cost dimension using a particle swarm optimization algorithm to obtain the dynamically updated cost function.

[0100] In one specific embodiment, the dynamic correction module 604 includes: in response to detecting a sudden obstacle, acquiring sudden obstacle information, the sudden obstacle information including the speed and position of the sudden obstacle; determining a predicted trajectory of the sudden obstacle based on the sudden obstacle information, and generating a dynamically feasible region based on the predicted trajectory of the sudden obstacle; correcting the initial lateral movement trajectory based on the dynamically feasible region and a cost function to obtain an intermediate lateral movement trajectory; determining whether the intermediate lateral movement trajectory satisfies the dynamic feasibility; if the dynamic feasibility is satisfied, using the intermediate lateral movement trajectory as the vehicle's lateral movement trajectory.

[0101] In one specific embodiment, the instruction generation module 605 includes: determining the desired lateral acceleration of the vehicle at the current time point based on the vehicle's lateral trajectory; acquiring the actual lateral acceleration actually performed by the vehicle at the current time point, and the steering system delay time; and generating vehicle control instructions based on the desired lateral acceleration, the actual lateral acceleration, and the steering system delay time.

[0102] Regarding the apparatus in the above embodiments, the specific manner in which each unit performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0103] Figure 7 This is a schematic diagram of the structure of a vehicle provided in an embodiment of this application.

[0104] For example, such as Figure 7As shown, the vehicle includes a memory 701 and a processor 702. The memory 701 stores executable program code 7011, and the processor 702 is used to call and execute the executable program code 7011 to perform a vehicle control method.

[0105] This embodiment can divide the vehicle into functional modules according to the above method example. For example, each function can be assigned to a separate module, or two or more functions can be integrated into one processing module. The integrated module can be implemented in hardware. It should be noted that the module division in this embodiment is illustrative and only represents one logical functional division. In actual implementation, there may be other division methods.

[0106] When each functional module is divided according to its corresponding function, the vehicle may include: The environmental perception module acquires spatiotemporal data of obstacles within the current control cycle and determines the predicted trajectory of the obstacles based on the spatiotemporal data. The trajectory modeling module predicts the trajectory based on the obstacles and models the lateral displacement of the vehicle at different future time points using high-order polynomials to obtain a variety of candidate lateral trajectories.

[0107] The trajectory planning module determines the vehicle's initial lateral movement trajectory from among the various candidate lateral movement trajectories, based on the current driving mode and a dynamically updated cost function.

[0108] The dynamic correction module, in response to the detection of a sudden obstacle, uses rolling time-domain planning to dynamically correct the initial lateral trajectory to obtain the vehicle's lateral trajectory.

[0109] The instruction generation module generates vehicle control instructions based on the vehicle's lateral movement trajectory, and controls the vehicle based on the vehicle control instructions so that the vehicle moves laterally according to the vehicle's lateral movement trajectory.

[0110] It should be noted that all relevant content of each step involved in the above method embodiments can be referenced from the functional description of the corresponding functional module, and will not be repeated here.

[0111] The vehicle provided in this embodiment is used to execute the above-described vehicle control method, and therefore can achieve the same effect as the above-described implementation method.

[0112] When using integrated units, the vehicle may include a processing module and a storage module. The processing module is used to control and manage the vehicle's actions. The storage module supports the vehicle in executing program code and data.

[0113] The processing module may be a processor or a controller, which can implement or execute various exemplary logic blocks, modules, and circuits as disclosed in this application. The processor may also be a combination of computing functions, such as a combination of one or more microprocessors, a combination of digital signal processing (DSP) and a microprocessor, etc., and the storage module may be a memory.

[0114] This embodiment also provides a computer-readable storage medium (including but not limited to disk storage, CD-ROM, optical storage, etc.) storing computer program code. When the computer program code is run on a computer, the computer executes the above-mentioned related method steps to implement the vehicle control method provided in the above embodiment.

[0115] This embodiment also provides a computer program product that, when run on a computer, causes the computer to perform the aforementioned steps to implement the vehicle control method provided in the above embodiment.

[0116] The beneficial effects of the above embodiments can be referred to the beneficial effects of the corresponding methods provided above, and will not be repeated here.

[0117] Through the above description of the embodiments, those skilled in the art will understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0118] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another apparatus, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0119] In the description of this disclosure, it should be understood that if the terms "upper", "lower", "front", "rear", "left" and "right" are used to indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the position or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this disclosure.

[0120] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0121] The above are merely embodiments of this disclosure and are not intended to limit the scope of this disclosure. Various modifications and variations can be made to this disclosure by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this disclosure should be included within the scope of the claims of this disclosure.

Claims

1. A vehicle control method, characterized in that, include: Acquire spatiotemporal data of obstacles within the current control cycle, and determine the predicted trajectory of obstacles based on the spatiotemporal data; Based on the obstacle-predicted trajectory, the lateral displacement of the vehicle at different future time points is modeled using a high-order polynomial to obtain a variety of candidate lateral trajectories. Based on the current driving mode and the dynamically updated cost function, the initial lateral movement trajectory of the vehicle is determined from the multiple candidate lateral movement trajectories. In response to the detection of a sudden obstacle, the initial lateral trajectory is dynamically corrected to obtain the vehicle's lateral trajectory; Based on the vehicle's lateral movement trajectory, a vehicle control command is generated, and the vehicle is controlled based on the vehicle control command so that the vehicle moves laterally according to the vehicle's lateral movement trajectory.

2. The method according to claim 1, characterized in that, The step of acquiring spatiotemporal data of obstacles within the current control cycle and determining the predicted trajectory of obstacles based on the spatiotemporal data specifically includes: The system acquires spatiotemporal data of all obstacles within a preset range of the vehicle using detection sensors pre-installed on the vehicle. The spatiotemporal data of all obstacles are input into the trajectory prediction model to obtain the predicted trajectories of all obstacles.

3. The method according to claim 1, characterized in that, The method for predicting trajectories based on obstacles involves modeling the lateral displacement of the vehicle at different future time points using high-order polynomials to obtain multiple candidate lateral trajectories, specifically including: Based on the predicted obstacle trajectory, the distance between each obstacle and the vehicle is determined; Obstacles with a distance greater than a preset threshold are removed to identify conflicting obstacles among all obstacles. Determine the predicted trajectory of the collision obstacle and the vehicle's current driving intention; Based on the predicted trajectory of the conflict obstacle and the current driving intention, a set of candidate trajectories for the vehicle is generated; By modeling the lateral displacement of vehicles in the candidate trajectory set at different future time points using high-order polynomials, a variety of candidate lateral displacement trajectories can be obtained.

4. The method according to claim 3, characterized in that, The process involves modeling the lateral displacement of vehicles in the candidate trajectory set at different future time points using high-order polynomials to obtain multiple candidate lateral displacement trajectories, specifically including: The lateral displacement of the vehicle is represented by a univariate high-order polynomial with time as the independent variable. Obtain the vehicle's current lateral position, current lateral velocity, and current lateral acceleration information; Determine the vehicle's desired lateral position at the end of the current control cycle; The current lateral position, current lateral velocity, current lateral acceleration information, desired lateral position, desired lateral velocity, and desired lateral acceleration are used as constraints on the univariate high-order polynomial. Wherein, the desired lateral velocity and the desired lateral acceleration are zero.

5. The method according to claim 1, characterized in that, The cost dimension of the cost function includes at least one of the following: time efficiency dimension, acceleration smoothness dimension, jerk comfort dimension, and target deviation distance dimension. Based on the predicted trajectory of the collision obstacle, the current driving mode, and the dynamically updated cost function, before outputting the vehicle's initial lateral movement trajectory from the multiple candidate lateral movement trajectories, the method further includes: Determine the vehicle's current driving mode; The weight range of each cost dimension in the cost function is determined based on the current driving mode. The weights of each cost dimension are dynamically optimized using the particle swarm optimization algorithm to obtain the dynamically updated cost function.

6. The method according to claim 1, characterized in that, In response to the detection of a sudden obstacle, the initial lateral trajectory is dynamically corrected using rolling time-domain programming to obtain the vehicle's lateral trajectory, specifically including: In response to the detection of a sudden obstacle, information about the sudden obstacle is acquired, including the speed and location of the sudden obstacle. Based on the information about the sudden obstacles, the predicted trajectory of the sudden obstacles is determined, and a dynamic feasible area is generated based on the predicted trajectory of the sudden obstacles. Based on the dynamic feasible region and the cost function, the initial lateral trajectory is corrected to obtain the intermediate lateral trajectory; Determine whether the intermediate lateral trajectory satisfies the dynamic feasibility requirement; If the aforementioned dynamic feasibility is satisfied, the intermediate lateral movement trajectory shall be taken as the vehicle's lateral movement trajectory.

7. The method according to claim 6, characterized in that, The determination of whether the intermediate lateral trajectory satisfies dynamic feasibility specifically includes: Obtain the lateral acceleration and lateral distance corresponding to each trajectory point in the intermediate lateral trajectory; Based on the lateral acceleration and the lateral distance, determine the maximum vehicle curvature corresponding to the intermediate lateral trajectory; When the maximum curvature of the vehicle is not higher than the preset curvature, the intermediate lateral movement trajectory satisfies the dynamic feasibility.

8. The method according to claim 1, characterized in that, The generation of vehicle control commands based on the vehicle's lateral trajectory specifically includes: Based on the vehicle's lateral trajectory, determine the vehicle's expected lateral acceleration at the current time point; Obtain the actual lateral acceleration of the vehicle at the current time point, as well as the steering system delay time; Based on the desired lateral acceleration, the actual lateral acceleration, and the steering system delay time, vehicle control commands are generated.

9. A vehicle, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the vehicle control method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing computer-executable instructions, characterized in that, The computer-executable instructions are configured to perform the vehicle control method as described in any one of claims 1 to 7.

Images