Path planning method based on vehicle model identification

Abstract

The invention provides a path planning method based on vehicle model identification, and aims to optimize lane changing decision and track generation of an automatic driving vehicle so as to improve driving safety and efficiency. The algorithm is composed of a vehicle model identification module, an acceleration performance evaluation module, a dynamic path planning module and a lane changing track generation module. The method is suitable for complex traffic environments including expressways, urban roads and congestion scenes, path planning is optimized by fully utilizing acceleration performance information of the target vehicle through a dynamic weight adjustment strategy, and the flexibility, intelligence and safety of an automatic driving system are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of autonomous driving technology, specifically relating to a path planning method based on vehicle model recognition, aiming to optimize lane-changing decisions and execution by autonomous vehicles. By identifying the brand and model of surrounding vehicles and combining their acceleration performance characteristics, the system can comprehensively analyze the impact of target vehicles on traffic flow, thereby improving the flexibility of lane-changing planning. This invention particularly focuses on lane-changing behavior in complex traffic scenarios, considering the differences in acceleration capabilities of different vehicle models. Through dynamic path planning and optimized trajectory generation, it achieves a balance between safety and efficiency for autonomous vehicles, significantly improving the intelligence level and operational effectiveness of the autonomous driving system. Background Technology

[0002] In recent years, with the rapid development of autonomous driving technology, path planning, as one of the core modules of autonomous driving systems, directly affects the vehicle's driving efficiency and overall safety. The quality of path planning not only relates to whether a vehicle can efficiently navigate complex traffic environments but also determines its risk management capabilities in dynamic scenarios. In current autonomous driving technologies, lane-change planning typically relies on the longitudinal distance and speed information of surrounding vehicles, making decisions by assessing the dynamic relationship between the two vehicles. However, this method ignores the significant differences in acceleration performance between different types of vehicles.

[0003] For example, due to design limitations, vans or minivans often have lower acceleration capabilities than cars or sports vehicles. This difference in acceleration performance directly affects their dynamic response capabilities on highways or congested roads. However, traditional path planning methods typically rely on fixed safety distances and speed difference thresholds, failing to fully consider the actual impact of vehicle type on lane-changing behavior, leading to conservative decision-making. Especially in complex traffic scenarios, the inability to effectively identify vehicles with low acceleration performance can cause a decrease in traffic flow efficiency, limiting the potential of autonomous vehicles in terms of flexibility and intelligence. Therefore, lane-changing planning algorithms optimized for vehicle type and acceleration performance differences are particularly important. This not only helps improve lane-changing efficiency but also better adapts to complex and changing traffic environments. Therefore, developing a path planning method that can identify vehicle models and integrate their acceleration performance to achieve more efficient and safer lane-changing decisions is of significant technical importance.

[0004] For example, in patent CN 118991771A "A design method for IVCPS lane-changing strategy based on traffic events", a lane-changing instruction can only be triggered after reaching a designated area.

[0005] For example, patent CN 117698726A, "Lane Changing Control Method, Device, On-board Terminal, and Vehicle for Vehicles," describes an invention that acquires the current vehicle state and environmental information of the target vehicle's surroundings. Based on this information, it evaluates the feasibility of actions of each candidate vehicle in the target vehicle's action space, obtains a reward value for each candidate vehicle's action, and controls the target vehicle to execute a lane-changing strategy based on these reward values. However, this lane-changing strategy only considers the action information of surrounding vehicles and does not take into account attributes such as the autonomous acceleration capabilities of surrounding vehicles.

[0006] In response, this invention proposes a path planning method based on vehicle model recognition, which integrates vehicle brand, model and acceleration performance to optimize lane-changing decisions of autonomous vehicles and improve the safety and driving efficiency of autonomous driving systems. Summary of the Invention

[0007] The path planning method based on vehicle model recognition of this invention achieves efficient and safe lane change planning through the coordinated work of four modules: vehicle model recognition, acceleration performance evaluation, dynamic path planning, and lane change trajectory generation. The following is a detailed description of each module of the technical solution:

[0008] 1. Vehicle model recognition module

[0009] Vehicle model recognition is the foundation of the entire path planning method. Its purpose is to identify the brand and model of surrounding vehicles through appearance images, obtain vehicle performance information, and use it to guide subsequent lane-changing decisions.

[0010] The vehicle model recognition module captures images of vehicles in the front and side lanes using an onboard camera. To ensure accuracy, the camera uses a high-definition wide-angle lens and is equipped with image stabilization to adapt to different road conditions. Simultaneously, the system is configured with frame rate parameters to ensure clear images are captured even at high speeds.

[0011] Computer vision techniques (such as deep learning algorithms) are used to extract vehicle exterior features, including brand logos, body shape, and lighting structure. The image processing workflow includes denoising, edge detection, feature extraction, and classification. The classification model outputs the vehicle's brand and specific model information based on a pre-trained vehicle model database.

[0012] When an image is blurry or unrecognizable, the system triggers an exception handling mechanism. For example, it may attempt to re-identify the vehicle using multi-frame image fusion technology, or classify unclassifiable vehicles as "unknown vehicles" and use the average performance of acceleration data for subsequent planning.

[0013] 2. Accelerated Performance Evaluation Module

[0014] After vehicle model identification is completed, the system queries the vehicle performance database based on the vehicle brand and model to evaluate the dynamic performance of the target vehicle (such as acceleration capability, top speed, etc.) and provide a reference for route planning.

[0015] The primary task of the acceleration performance evaluation module is to quantitatively analyze the dynamic performance of surrounding vehicles, providing crucial decision-making support for path planning. Through comprehensive analysis of vehicle acceleration capabilities, speed, and operating environment, the system can identify the potential impact of target vehicles on lane-changing behavior and optimize lane-changing priorities and path planning strategies.

[0016] The system relies on a performance database that covers the brands and models of mainstream vehicles on the market, along with their dynamic performance data. The database fields include vehicle model, 0-100km / h acceleration time, maximum speed, vehicle curb weight, and powertrain type. Data sources include publicly available technical documentation from vehicle manufacturers, historical operational data from fleet management platforms, and machine learning models that interpolate and predict based on similar vehicle models. Real-time updates and expansion of the performance database ensure the system's adaptability to new and specialized vehicle models.

[0017] After obtaining the brand and model of the target vehicle, the system extracts relevant dynamic performance parameters from the performance database and calculates the vehicle's average acceleration 'a'. avg This quantifies its acceleration capability. Simultaneously, the system generates an acceleration performance score S by comparing its acceleration capability with that of a reference vehicle. a The performance score reflects the relative level of the target vehicle's dynamic performance. A lower score indicates weaker acceleration, thus increasing its priority in lane-change planning. To adapt to different traffic scenarios, the system dynamically adjusts the performance score weights based on the target vehicle's environment. For example, on highways, the system gives higher weight to acceleration to avoid reducing traffic flow efficiency due to low-performance vehicles; while on urban roads, the impact of acceleration performance is relatively reduced.

[0018] The module combines real-time speed information to analyze the relative speed Δv between the target vehicle and the current vehicle. When the relative speed is high, even if the target vehicle has low acceleration performance, the system will prioritize avoiding potentially risky vehicles; conversely, when the speed difference is small, the impact of acceleration performance is amplified. Furthermore, in special scenarios such as slopes or slippery surfaces, the system adjusts the acceleration performance score to prioritize safety. For example, in uphill scenarios, vehicles with weaker acceleration will significantly increase their impact on lane-changing planning, while in rainy or snowy weather, the acceleration performance of all vehicles will decrease, so the system reduces the weight of the performance score to prioritize safe driving.

[0019] The acceleration performance evaluation module outputs data such as performance scores, speed differences, and dynamic weights. This data serves as input variables for the path planning module, supporting the generation of efficient and safe lane-changing paths. By introducing this module, the system can accurately identify and address the impact of vehicles with low acceleration performance (such as pickup trucks or vans) on path planning, thereby improving the driving efficiency and safety of autonomous vehicles in complex traffic environments.

[0020] 3. Dynamic Path Planning Module

[0021] The dynamic route planning module is responsible for generating the optimal lane-changing route based on real-time traffic conditions and vehicle dynamic status, ensuring driving safety and efficiency. Its main functions include data acquisition and fusion, lane-changing condition judgment, route priority evaluation, and route generation and optimization. Through comprehensive analysis of the environment and vehicle status, it enables flexible lane-changing decisions.

[0022] Before planning a route, the system first collects data on the surrounding traffic environment using multiple sensors. Onboard cameras provide vehicle position information for the vehicles ahead and in adjacent lanes, while also acquiring road geometry features such as lane width and curve radius. LiDAR and millimeter-wave radar detect the speed and acceleration of surrounding vehicles in real time and measure the relative distances between vehicles. This multi-source data is comprehensively processed using fusion techniques such as Kalman filtering to eliminate noise and redundant information, forming a dynamic environmental model that includes vehicle position, speed, acceleration, and road structure. This provides accurate input for subsequent lane-changing condition determination.

[0023] Determining lane-changing conditions is a crucial step in path planning. The system needs to comprehensively evaluate longitudinal safety distance, relative speed, and the availability of the target lane to ensure the rationality of lane-changing actions. The longitudinal safety distance is calculated based on relative speed v. rel System reaction time t reaction and minimum safe distance d min If the actual distance d between this vehicle and the vehicle in the target lane... actual >d safe If the speed of a vehicle in the target lane is significantly lower than the vehicle's speed (i.e., Δv is large and negative), the system will prioritize lane changing to avoid impacting driving efficiency due to low-speed vehicles. Conversely, when Δv is small, the need for lane changing decreases accordingly. Simultaneously, the system assesses the availability of the target lane based on road environment characteristics (such as whether it is on a curve or near an exit lane). For example, when on a curve or near an exit, the system may reduce lane changing priority to minimize operational risks.

[0024] After the lane-changing conditions are met, the system generates multiple possible paths and prioritizes them to select the optimal solution. Path priority evaluation is based on a multi-objective optimization model, comprehensively considering longitudinal safety distance, relative speed, and the acceleration performance weight of the target vehicle. For example, on highways, the system assigns higher weights to acceleration performance and speed difference, while on urban roads, the weight of safety distance is emphasized.

[0025] The system generates lane-change paths based on priority evaluation results and ensures the smoothness and feasibility of the paths through trajectory optimization. The paths are generated using cubic spline interpolation, with the starting point being the vehicle's current state (position, speed, and acceleration) and the ending point being the safe position in the target lane. Intermediate control points are dynamically adjusted based on the status of surrounding vehicles and road characteristics to ensure the path avoids obstacles and meets vehicle dynamics constraints. During optimization, the system improves ride smoothness by minimizing path curvature changes, avoiding the impact of sharp turns or discontinuous trajectories on vehicle stability and passenger comfort.

[0026] The module also has real-time adjustment capabilities to cope with dynamically changing traffic environments. During lane change execution, if a new vehicle suddenly appears in the target lane or the vehicle ahead slows down, the system can reassess the path based on the latest sensor data and adjust trajectory parameters in real time, canceling the lane change operation if necessary to avoid potential risks.

[0027] 4. Lane Change Trajectory Generation Module

[0028] The lane change trajectory generation module is responsible for converting the planned lane change path into a specific driving trajectory, ensuring a smooth and safe lane change operation.

[0029] Based on the lane-changing path provided by the planning module, the system generates a specific lane-changing trajectory, including the start and end points of the lane change, as well as the smoothness of the trajectory. The start and end points of the lane change are dynamically adjusted according to the positions and speeds of surrounding vehicles and road conditions.

[0030] To avoid vehicle instability or passenger discomfort caused by sharp turns, the system optimizes lane-changing paths using a trajectory smoothing algorithm. For example, the curvature change of the lane-changing trajectory must be continuous, avoiding abrupt changes.

[0031] During lane change execution, the module dynamically adjusts its trajectory based on the latest sensor data. For example, if a new vehicle suddenly appears in an adjacent lane or a target vehicle decelerates, the system will immediately correct its trajectory and cancel the lane change operation if necessary.

[0032] When generating a trajectory, vehicle dynamics constraints, such as acceleration and steering angle rate, must be met to ensure that the trajectory is feasible and does not pose a challenge to the vehicle's performance.

[0033] The beneficial effects of this invention are as follows: The path planning method based on vehicle model recognition provided by this invention is designed for urban roads. By combining data reported by different types of vehicles and fully considering vehicle models and their acceleration capabilities, it can quickly and effectively optimize lane-changing decisions for intelligent unmanned vehicles on urban roads. While ensuring vehicle safety, it effectively shortens the waiting time of vehicles on the road, improves the traffic efficiency of urban roads, and enhances environmental and economic benefits. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 A flowchart of a path planning method based on vehicle model recognition;

[0036] Figure 2 A schematic diagram of the module structure of the path planning method based on vehicle model recognition. Detailed Implementation

[0037] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and, together with the embodiments of the present invention, serve to illustrate the principles of the present invention.

[0038] This invention utilizes four core modules: vehicle model recognition, acceleration performance evaluation, dynamic path planning, and lane-change trajectory generation. Figure 1 As shown, this enables efficient path planning for autonomous vehicles, especially flexible lane changing in complex traffic scenarios. The workflow of each module is explained in detail below:

[0039] 1. Vehicle Model Identification Module S101

[0040] The goal of the vehicle model recognition module is to identify the brand and model of surrounding vehicles through image processing technology, providing performance data support for subsequent lane change planning.

[0041] (1) Data acquisition:

[0042] The system utilizes an onboard high-definition camera to capture images of vehicles ahead and in adjacent lanes. The camera's sampling frame rate is set to 20-30fps to meet the demands of high-speed scenarios. Image data is transmitted in real-time to a central processing unit for processing.

[0043] (2) Feature extraction and classification:

[0044] Vehicle appearance features f are extracted using a deep learning model (such as ResNet) to classify brand logos, body shape, etc. The model outputs the classification probability of vehicle brand and model through equation (1):

[0045]

[0046] Where i represents the vehicle category, w i and b i Here, represents the model weights and biases, respectively, and N is the total number of classification categories. The vehicle model y is selected based on the principle of maximizing probability. * .

[0047] (3) Exception handling:

[0048] If an image cannot be classified or is identified as an "unknown vehicle", the system uses multi-frame fusion or default average acceleration performance data for subsequent processing.

[0049] 2. Accelerated Performance Evaluation Module S102

[0050] After identifying the vehicle brand and model, the system queries the vehicle performance database to obtain dynamic performance data related to acceleration, providing parameter support for route planning.

[0051] (1) Accelerating performance database:

[0052] The database contains vehicle model y and 0-100km / h acceleration time t. 0-100 Maximum vehicle speed v max Key performance data:

[0053] D={(y,t 0-100 ,v max (2)

[0054] (2) Performance scoring and comparison:

[0055] Based on the identified model y * Extract the acceleration capability of the target vehicle. y *:

[0056]

[0057] Compared to a reference vehicle (such as a regular sedan), the acceleration performance is a ref Calculate performance weights:

[0058]

[0059] Performance weight w a The smaller the value, the lower the vehicle's acceleration capability, and the greater the impact on lane change planning priority.

[0060] (3) Dynamic adjustment of speed difference:

[0061] Combined with the real-time speed v of the target vehicle env And the speed of this car v ego Calculate the speed difference Δv:

[0062] Δv=v env -v ego (5)

[0063] When the speed difference Δv is large, the performance impact is further amplified, thus adjusting the lane-changing strategy.

[0064] 3. Dynamic Path Planning Module S103

[0065] The dynamic path planning module comprehensively evaluates lane-changing conditions based on the status information of the vehicle and surrounding vehicles, and generates a planned path.

[0066] (1) Lane changing condition judgment:

[0067] The system determines whether a lane change meets the following conditions:

[0068] d long >d safe (6)

[0069] Where, d long This is the longitudinal safety distance. d safe =v rel .t reaction +d min This represents the safe distance determined by relative speed and reaction time. Furthermore, when the vehicle speed difference Δv < 0 and |Δv| is large, lane changing is preferred.

[0070] (2) Lane change priority assessment:

[0071] By combining longitudinal distance, safety, and performance weights, a lane-changing priority score P is defined. lane :

[0072]

[0073] Here, α1, α2, and α3 are weighting parameters. Lane-changing options with higher priority will be used.

[0074] (3) Path generation:

[0075] The path planning uses a reference path-based search algorithm. After generating an initial path, the final path is generated through trajectory optimization.

[0076]

[0077] Where k(s) is the path curvature, v(s) is the path velocity, and λ is the smoothing factor. By solving the optimal control problem, a smooth path that satisfies the vehicle dynamic constraints is generated.

[0078] 4. Lane change trajectory generation module S104

[0079] The lane change trajectory generation module transforms the planned path into an executable trajectory for the vehicle, ensuring path smoothness and dynamic feasibility.

[0080] (1) Trajectory representation:

[0081] The lane-changing trajectory is represented using cubic splines:

[0082]

[0083] coefficient a i b i Solve based on boundary conditions (starting point, ending point, and derivative).

[0084] (2) Trajectory optimization:

[0085] The following objective function is solved through constraint optimization to ensure that the trajectory is collision-free, smooth, and satisfies dynamic constraints:

[0086]

[0087] Where β is the speed deviation penalty coefficient and T is the lane change time.

[0088] (3) Real-time adjustment:

[0089] If the environmental conditions change during trajectory execution (such as the target vehicle slowing down), the system will dynamically adjust the trajectory parameters, or even terminate the lane change to ensure safety.

[0090] Through the coordinated operation of the above modules, this invention achieves intelligent identification of different types of vehicles and performance-based optimized path planning, thereby improving the driving efficiency and safety of autonomous vehicles in complex traffic scenarios.

[0091] Figure 2 This is a schematic diagram of the module structure of a path planning method based on vehicle model recognition provided by the present invention, as shown below. Figure 2 As shown, the path planning system based on vehicle model recognition constructed by the above method mainly includes a vehicle model recognition module; an acceleration performance evaluation module; a dynamic path planning module and a lane change trajectory generation module.

[0092] Vehicle model recognition module 201: This module uses an onboard camera and deep learning algorithms to identify the brand and model of surrounding vehicles. Its core task is to acquire the basic characteristics of the target vehicle, providing data support for subsequent acceleration performance evaluation and ensuring that the dynamic performance differences between different vehicle models are fully considered during path planning.

[0093] Acceleration Performance Evaluation Module 202: Based on the vehicle model, this module extracts parameters such as acceleration time and maximum speed from the performance database and combines them with the target vehicle's real-time speed to calculate its acceleration performance score and relative speed. This module quantifies the target vehicle's impact on lane-changing behavior, helping the dynamic path planning module make better decisions.

[0094] Dynamic path planning module 203: Combining environmental perception data (such as vehicle position, speed, acceleration, and road characteristics), it determines whether a lane change is necessary and plans an optimal path while meeting safety and efficiency requirements. This module is responsible for determining the feasibility and priority of lane-changing behavior from a macroscopic perspective;

[0095] Lane change trajectory generation module 204: Based on the dynamic path planning results, this module generates a specific executable trajectory for the vehicle. By optimizing curvature changes, path smoothness, and dynamic constraints, this module ensures the physical feasibility and driving comfort of the lane change trajectory.

Claims

1. A path planning method based on vehicle model recognition, characterized in that, The algorithm includes the following steps: (1) Vehicle model identification The vehicle captures exterior images of surrounding vehicles using an in-vehicle camera, and uses a deep learning model to identify the brand and model of the target vehicle. It then extracts the acceleration performance parameters of the target vehicle, including acceleration time and maximum speed, from a pre-set vehicle performance database. (2) Acceleration performance evaluation Based on the identified vehicle brand and model, the average acceleration capability of the target vehicle is calculated, and combined with the real-time speed difference between the current vehicle and the target vehicle, an acceleration performance score is generated as the basis for path planning priority calculation. (3) Dynamic path planning By combining surrounding environment perception data and target vehicle performance data, the longitudinal distance, relative speed and road characteristics between the vehicle and the target vehicle are comprehensively analyzed to determine whether a lane change is necessary, and the optimal path is planned under the condition of meeting safety requirements. (4) Lane change trajectory generation Based on the planned path, specific lane-changing trajectories are generated, curvature changes are optimized, and vehicle dynamics constraints are considered to ensure the smoothness and executability of the trajectory.

2. The path planning method based on vehicle model recognition according to claim 1, characterized in that, In the vehicle model recognition step, a deep learning model is used to extract features from the images of surrounding vehicles, specifically extracting significant appearance features such as vehicle brand logos, body outlines, headlight structure, and color. The extracted feature vectors are then input into the trained classification model, which outputs the probability distribution of vehicle brands and models. The brand and model with the highest probability are selected as the recognition result. Furthermore, the system extracts the corresponding dynamic performance parameters from the performance database based on the identified vehicle brand and model information, including acceleration capability, maximum speed, and curb weight, for subsequent path planning.

3. The path planning method based on vehicle model recognition according to claim 1, characterized in that, The acceleration performance evaluation step includes the following steps for calculating the acceleration performance score of the target vehicle: (1) Extract the 0-100km / h acceleration time t of the target vehicle from the vehicle performance database. 0-100 and maximum vehicle speed v max Calculate its average acceleration (2) The average acceleration a of the target vehicle avg The average acceleration a of the reference vehicle ref The comparison yields an acceleration performance score. (3) Based on the acceleration performance score S a This assigns dynamic weights to the subsequent path planning module to adjust the degree of influence of the target vehicle on the lane-changing decision.

4. The path planning method based on vehicle model recognition according to claim 1, characterized in that, In the dynamic path planning step, the path priority score P is calculated based on a multi-objective optimization model. lane : Where, d long d represents the longitudinal distance between this vehicle and the vehicle in the target lane. ref For reference longitudinal distance; ω a The acceleration performance weight of the target vehicle reflects the impact of its dynamic performance on lane-changing decisions; Δv is the relative speed, representing the speed difference between the current vehicle and the vehicle in the target lane; v ego The speed is α1; α2, α3 are weighted parameters dynamically adjusted according to the traffic scenario. Paths with higher scores indicate higher overall safety and efficiency and are prioritized as lane-changing routes.

5. The path planning method based on vehicle model recognition according to claim 1, characterized in that, The lane-changing trajectory generation step uses cubic spline interpolation to generate the lane-changing trajectory, specifically including the following steps: (1) Based on the starting position (including current position, velocity and acceleration) and target position provided by the dynamic path planning module, set the starting point, ending point and several control points of the trajectory; (2) Calculate the curve equation of each trajectory segment using cubic spline interpolation to ensure the continuity of the trajectory in terms of position, velocity and acceleration; (3) The location of the control points is dynamically adjusted according to the surrounding vehicle status and road geometry to avoid collision risks and adapt to complex traffic environments; (4) The output lane change trajectory is smooth and meets the vehicle's motion constraints, which facilitates precise execution by autonomous vehicles.

6. The path planning method based on vehicle model recognition according to claim 1, characterized in that, In the lane-changing trajectory generation step, the smoothness and comfort of the trajectory are ensured by optimizing the curvature change and length of the lane-changing path, specifically including: (1) In the generated lane-changing trajectory, minimize the curvature change and avoid sharp turns or curvature discontinuities; (2) Consider vehicle dynamics constraints, including maximum steering angle rate, minimum turning radius and maximum acceleration, to ensure the physical executability of the trajectory; (3) Optimize the total path length and lane-changing time to improve execution efficiency while ensuring safety; (4) If the environment changes dynamically (such as new vehicles or obstacles appearing in the target lane), the trajectory parameters are updated in real time and the lane-changing trajectory is regenerated to adapt to the new traffic conditions.

7. The path planning method based on vehicle model recognition according to claim 1, characterized in that, The dynamic path planning module and lane change trajectory generation module have real-time adjustment capabilities. When the target lane in the environment undergoes sudden changes (such as a vehicle suddenly entering, a significant change in speed, or a deterioration in road conditions), the system can adjust the lane change strategy in the following ways: (1) Dynamically reassess lane-changing conditions, comprehensively analyze longitudinal distance, relative speed and acceleration performance weights, and update path priority scores; (2) Cancel the lane-changing decision if necessary, maintain the current lane and avoid potential collision risks; (3) Optimize the generated lane-changing trajectory in real time to ensure that the trajectory still meets the safety and feasibility requirements of the current traffic environment.

8. The path planning method based on vehicle model recognition according to claim 1, characterized in that, This algorithm is applicable to complex and diverse traffic scenarios, including but not limited to highways, urban roads, and multi-lane traffic environments. The system dynamically adjusts the weight parameters α1, α2, α3 and the safe distance threshold d. safe Optimize according to the characteristics of different scenarios: (1) In the highway scenario, the system prioritizes the acceleration performance and speed difference of the target vehicle to ensure efficient driving; (2) In urban roads, more attention should be paid to longitudinal safety distance and target lane availability, and priority should be given to ensuring safety in low-speed environments; (3) In congested environments or special scenarios (such as ramps or rainy / snowy weather), dynamically reduce the weight of acceleration performance and increase the priority of safety constraints to adapt to complex traffic needs.

Images