Decision-making method and device of unmanned vehicle and vehicle mixed driving control system

Abstract

The invention provides an unmanned vehicle decision-making method and device and a vehicle mixed driving control system. The method comprises the following steps: detecting an obstacle in the driving direction of an unmanned vehicle; under the condition that the key obstacle is detected, identifying the type of the key obstacle; wherein the key obstacles comprise obstacles possibly interacting with the unmanned vehicle; generating a sampling path in a preset area range in the driving direction of the unmanned vehicle, performing collision detection on the sampling path based on the key obstacle of the first preset type, and screening out a first sampling path; determining an interaction intention between the unmanned vehicle and the key obstacle of the second preset type according to the first dynamic intention of the key obstacle of the second preset type and the second dynamic intention of the unmanned vehicle when the unmanned vehicle runs in the first sampling path; and evaluating the interaction intention, and generating a driving strategy of the unmanned vehicle according to an evaluation result. The problem of safe and efficient decision making of the unmanned vehicle facing various obstacles in a complex dynamic traffic environment is solved.

Description

Technical Field

[0001] The present disclosure relates to the field of driverless technologies, and in particular, to a decision-making method, device, and vehicle mixed-traffic control system for driverless vehicles. Background Art

[0002] In the field of intelligent transportation, the development of driverless technologies is of great significance for improving traffic safety and passing efficiency. For driverless scenarios, safety is the basic requirement for protecting public life and property, while efficiency directly affects the practicality and promotion value of the technologies.

[0003] Currently, driverless operation scenarios generally face complexity challenges. In terms of road conditions, in addition to urban structured roads, there are also a large number of unstructured or semi-structured roads, and some sections lack clear lane line markings, making it difficult for vehicles to regulate their driving trajectories through traditional markings. At the same time, traffic intersections have diverse forms, and situations such as irregular roundabouts and multi-way intersections are common. The open road environment further increases the uncertainty of driving decisions.

[0004] In addition, during actual operation, driverless vehicles need to interact frequently with a large number of manned vehicles, non-motor vehicles, and pedestrians. For example, in scenarios such as logistics transportation and public transportation, there are high-frequency contacts between driverless vehicles and auxiliary operation equipment and social vehicles, and the interaction behaviors are complex and variable. Against this background, how to improve the intelligence of the interaction between autonomous driving vehicles and other dynamic traffic participants while ensuring safety has become the core requirement for the development of driverless technologies. This not only requires more advanced perception technologies to accurately identify various targets, but also relies on efficient decision-making algorithms to achieve safe and efficient driving in complex scenarios. Summary of the Invention

[0005] Embodiments of the present disclosure provide a decision-making method, device, and vehicle mixed-traffic control system for driverless vehicles to solve the related problems existing in the existing technical solutions.

[0006] Based on the above problems, in a first aspect, a decision-making method for a driverless vehicle is provided, including: Detect obstacles in the driving direction of the driverless vehicle; When a key obstacle is detected, identify the type of the key obstacle; where the key obstacle includes an obstacle that may interact with the driverless vehicle; Generate a sampling path within a preset area range in the driving direction of the driverless vehicle, and perform a collision detection on the sampling path based on a first preset type of key obstacle, and filter out a first sampling path; Determine the interaction intention between the driverless vehicle and the key obstacle of the second preset type according to the first dynamic intention of the key obstacle of the second preset type and the second dynamic intention of the driverless vehicle when driving on the first sampling path; Evaluate the interaction intention and generate a driving strategy for the driverless vehicle according to the evaluation result.

[0007] Combined with the first aspect, in a possible implementation manner, the obstacles include: non-interaction obstacles and key obstacles; the key obstacles include: key static obstacles, abnormal obstacles, and key dynamic obstacles; the key static obstacles include: constant static obstacles and active static obstacles.

[0008] Combined with the first aspect, in a possible implementation manner, generating a sampling path within a preset area range in the driving direction of the driverless vehicle and performing a collision detection on the sampling path based on the key obstacle of the first preset type, and screening out the first sampling path, includes: Generate sampling points within a preset area range in the driving direction of the driverless vehicle, and perform a first collision detection on the sampling points based on the key obstacle of the first preset type, and screen out the first sampling points; Generate multiple sampling paths based on the first sampling points, and perform a second collision detection on the multiple sampling paths based on the key obstacle of the first preset type, and screen out the first sampling path.

[0009] Combined with the first aspect, in a possible implementation manner, the preset area range includes: a longitudinal length and a lateral width; Generating sampling points within a preset area range in the driving direction of the driverless vehicle, includes: Determine the longitudinal sampling interval on the longitudinal length based on the longitudinal remaining length of the predicted trajectory of the driverless vehicle and / or the longitudinal position of the key obstacle; Generate longitudinal sampling points on the longitudinal length in the driving direction of the driverless vehicle according to the longitudinal sampling interval; Determine the lateral sampling interval on the lateral width corresponding to each longitudinal sampling point based on the lateral position of the key obstacle and the lateral width of the drivable space in the lateral direction of each longitudinal sampling point; Generate lateral sampling points in the lateral direction of the position where each longitudinal sampling point is located according to the corresponding lateral sampling interval; Wherein, the longitudinal length is determined based on the driving speed of the driverless vehicle; the lateral width is determined based on the lateral drivable boundary of the driverless vehicle, and satisfies: the closer to the driverless vehicle, the narrower the lateral width.

[0010] In combination with the first aspect, in a possible implementation, the key obstacles of the first preset type include abnormal obstacles and key static obstacles; The first collision detection of the sampling points based on the key obstacles of the first preset type includes: Determine whether there is an overlap between the sampling point and the key obstacles of the first preset type. If there is an overlap, remove the sampling point to obtain the first sampling points that do not overlap with the key obstacles of the first preset type; The second collision detection of the multiple sampling paths based on the key obstacles of the first preset type to screen out the first sampling paths includes: Determine whether there is an intersection between the multiple sampling paths and the key obstacles of the first preset type. If there is an intersection, remove the sampling path to obtain the first sampling paths that do not intersect with the key obstacles of the first preset type.

[0011] In combination with the first aspect, in a possible implementation, after obtaining the first sampling points, it further includes: Exclude the first sampling points that exceed the steering ability of the driverless vehicle from the first sampling points to obtain the first sampling points after re-screening; Generate multiple sampling paths based on the first sampling points, including: Generate multiple sampling paths based on the first sampling points after re-screening.

[0012] In combination with the first aspect, in a possible implementation, the key obstacles of the second preset type include key dynamic obstacles; Determine the interaction intention between the driverless vehicle and the key obstacles of the second preset type according to the first dynamic intention of the key obstacles of the second preset type and the second dynamic intention of the driverless vehicle when driving on the first sampling path, including: Obtain the predicted trajectory of the key dynamic obstacle, and sample various accelerations of the key dynamic obstacle when driving on the predicted trajectory to obtain the first dynamic intention; For each first sampling path, sample various accelerations of the driverless vehicle when driving on the first sampling path to obtain the second dynamic intention; Combine the first dynamic intention and the second dynamic intention with the first sampling path to determine the interaction intention of the driverless vehicle to avoid interacting with the key dynamic obstacle during the driving process on the first sampling path.

[0013] In combination with the first aspect, in a possible implementation, evaluate the interaction intention and generate the driving strategy of the driverless vehicle according to the evaluation result, including: Evaluate the preset metrics for each interaction intention; wherein, the preset metrics include at least one of the following: safety, anthropomorphism, and passability; According to the evaluation results of each interaction intention, perform an evaluation and sorting on the interaction intention; Based on the multi-intention requirements, classify and code the sorted interaction intentions, and eliminate the duplicate intentions with lower rankings for each classification code; Generate the driving strategy of the driverless vehicle based on the remaining interaction intentions after eliminating duplicate intentions and the evaluation and sorting results of the remaining interaction intentions.

[0014] Combined with the first aspect, in a possible implementation manner, before determining the interaction intention between the driverless vehicle and the key obstacle of the second preset type when the driverless vehicle travels on the first sampling path according to the first dynamic intention of the key obstacle of the second preset type and the second dynamic intention of the driverless vehicle, it further includes: Evaluate and sort the first sampling path based on at least one of the path form change variance, distance from the reference line, distance from the active static obstacle, and end point distance of the first sampling path, and obtain the first sampling paths with the top preset number of rankings; Wherein, the smaller the path form change variance, the higher the corresponding evaluation value; the smaller the distance from the reference line, the higher the corresponding evaluation value; the larger the distance from the active static obstacle, the higher the corresponding evaluation value; the smaller the end point distance, the higher the corresponding evaluation value; The determining the interaction intention between the driverless vehicle and the key obstacle of the second preset type according to the first dynamic intention of the key obstacle of the second preset type and the second dynamic intention of the driverless vehicle when the driverless vehicle travels on the first sampling path includes: Determine the interaction intention between the driverless vehicle and the key obstacle of the second preset type according to the first dynamic intention of the key obstacle of the second preset type and the second dynamic intention of the driverless vehicle when the driverless vehicle travels on the first sampling paths with the top preset number of rankings.

[0015] Combined with the first aspect, in a possible implementation manner, the performing type recognition on the key obstacle when detecting the key obstacle includes: When detecting the key obstacle, perform type recognition on the obstacle according to the reference line projection of the key obstacle; The method further includes: When the obstacle is an abnormal obstacle, inflate the obstacle to form an abnormal driving area as the obstacle corresponding area; When the obstacle is an obstacle that normally interacts with the driverless vehicle, obtain the proximal specified area of the predicted trajectory of the obstacle as the obstacle corresponding area; Construct an SL map for the obstacle corresponding area based on different types of obstacles.

[0016] In a second aspect, a decision-making device for a driverless vehicle is provided, including: An obstacle detection module for detecting obstacles in the driving direction of the vehicle. An obstacle recognition module for identifying the type of the key obstacle when the key obstacle is detected; wherein, the key obstacle includes an obstacle that may interact with the vehicle. A collision test module for generating a sampling path within a preset area range in the driving direction of the vehicle, and performing a collision detection on the sampling path based on the key obstacle of the first preset type to screen out the first sampling path. An interaction intention module for determining the interaction intention between the vehicle and the key obstacle of the second preset type according to the first dynamic intention of the key obstacle of the second preset type and the second dynamic intention of the vehicle when driving on the first sampling path. A strategy generation module for evaluating the interaction intention and generating a driving strategy for the vehicle according to the evaluation result.

[0017] In a third aspect, a vehicle mixed traffic control system is provided, including a decision-making device for a vehicle as described in the second aspect or any possible implementation manner combining the second aspect.

[0018] The beneficial effects of the embodiments of the present disclosure include: Embodiments of the present disclosure provide a decision-making method, device, and vehicle mixed-traffic control system for driverless vehicles, which are applied to the interaction of driverless vehicles with a large number of other vehicles and the response to irregular obstacles in various complex scenarios. Specifically, the decision-making method includes: First, detect obstacles in the driving direction of the driverless vehicle, and obtain information such as their positions and speeds, providing a data basis for subsequent decision-making. Next, identify the types of key obstacles that may cause interactions, distinguish dynamic objects such as pedestrians and vehicles from static objects such as guardrails, and handle different obstacles in different ways. Then, generate sampling paths in a preset area, perform collision detection based on the first preset type of key obstacles, and filter out the first sampling paths without direct collision risks, reducing the decision-making complexity. Further, according to the first dynamic intention of the second preset type of key obstacles and the second dynamic intention of the driverless vehicle when driving on the first sampling path, determine the interaction intention of the driverless vehicle with respect to the second preset type of obstacles, ensuring the safety and accuracy of the interaction intention from two aspects. Finally, conduct a multi-dimensional evaluation of the interaction intention to generate the final driving strategy. In summary, this method solves the problem of safe and efficient decision-making for driverless vehicles facing obstacles in a dynamic traffic environment, hierarchically processes different types of obstacles, combines intention analysis and multi-dimensional evaluation, reduces accident risks, and realizes the driving decision-making of driverless vehicles under complex road conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is one of the flowcharts of a decision-making method for a driverless vehicle provided by an embodiment of the present disclosure; Figure 2 is a schematic diagram of various interaction types provided by an embodiment of the present disclosure; Figure 3 is a schematic diagram of a collision test provided by an embodiment of the present disclosure; Figure 4 is one of the schematic diagrams of acceleration sampling provided by an embodiment of the present disclosure; Figure 5 is another schematic diagram of acceleration sampling provided by an embodiment of the present disclosure; Figure 6 is a flowchart of obstacle type identification provided by an embodiment of the present disclosure; Figure 7 is another flowchart of a decision-making method for a driverless vehicle provided by an embodiment of the present disclosure; Figure 8 is a schematic diagram of a decision-making device for a driverless vehicle provided by an embodiment of the present disclosure. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] Embodiments of the present disclosure provide a decision-making method, device, and vehicle mixed traffic control system for driverless vehicles. The preferred embodiments of the present disclosure will be described below with reference to the accompanying drawings of the specification. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present disclosure and are not used to limit the present disclosure. And without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other.

[0021] Embodiments of the present disclosure provide a decision-making method for driverless vehicles, as Figure 1 shown, which can be implemented as: S101. Detect obstacles in the driving direction of the driverless vehicle; S102. When a key obstacle is detected, identify the type of the key obstacle; wherein, the key obstacle includes an obstacle that may interact with the driverless vehicle; S103. Generate a sampling path within a preset area range in the driving direction of the driverless vehicle, and perform collision detection on the sampling path based on the key obstacles of the first preset type, and filter out the first sampling path; S104. Determine the interaction intention between the driverless vehicle and the key obstacles of the second preset type according to the first dynamic intention of the key obstacles of the second preset type and the second dynamic intention of the driverless vehicle when driving on the first sampling path; S105. Evaluate the interaction intention and generate a driving strategy for the driverless vehicle according to the evaluation result.

[0022] When a driverless vehicle is driving in a complex road environment, it is necessary to generate accurate driving decisions in real time to ensure safety and traffic efficiency. For this reason, the present disclosure provides a decision-making method for driverless vehicles, which specifically processes different types of obstacles and generates a reasonable driving strategy.

[0023] In the embodiments of the present disclosure, first, obstacles in the driving direction of the driverless vehicle are detected. This step uses sensors such as lidar, cameras, and millimeter-wave radars deployed on the driverless vehicle to obtain basic information such as the position, speed, and size of the obstacles existing in its driving direction, laying a data foundation for subsequent determination of the type of obstacles and generation of the final driving strategy of the driverless vehicle.

[0024] Then, it is judged whether the detected obstacle is likely to interact with the driverless vehicle (ego vehicle), and the obstacle that may interact with the driverless vehicle is a key obstacle. After detecting a key obstacle that may interact with the driverless vehicle, its type is identified. Among them, techniques such as image recognition and deep learning algorithms can be used to identify the type of obstacle, which is not limited here.

[0025] Next, multiple sampling paths are generated within a pre-set area in the direction of the autonomous vehicle's travel. Collision detection is performed against key obstacles of a first pre-set type, eliminating paths with collision risks and selecting theoretically safe first sampling paths. This narrows the scope of subsequent decisions and improves decision-making efficiency. Key obstacles of the first pre-set type are described in detail later in this article and include static obstacles such as stationary vehicles and fixed isolation piers, so they will not be further elaborated here.

[0026] Afterwards, based on the dynamic intentions of the second preset type of key obstacles, such as pedestrians suddenly crossing the road and vehicles changing lanes, combined with the intentions of the unmanned vehicle when driving on the first sampling path, the interaction intentions of both parties are determined, and it is clarified whether the unmanned vehicle needs to slow down to give way or speed up to pass.

[0027] Finally, the interaction intention is evaluated from the dimensions of safety, traffic efficiency, comfort, etc., and the evaluation results are converted into a series of specific driving strategies with execution priorities.

[0028] In summary, the autonomous vehicle decision-making method provided by this disclosure not only achieves basic collision avoidance by layering different types of obstacles, combining dynamic intent analysis with multi-dimensional assessment, but also improves the rationality and diversity of decision-making in complex interactive scenarios. Compared to traditional decision-making methods, this method classifies key obstacles and optimizes paths and strategies in a step-by-step manner, effectively reducing the risk of traffic accidents and improving the efficiency and ride comfort of autonomous vehicles in complex road conditions.

[0029] In another embodiment provided by the present disclosure, obstacles include: non-interactive obstacles and key obstacles; key obstacles include: key static obstacles, abnormal obstacles and key dynamic obstacles; key static obstacles include: normal static obstacles and active static obstacles.

[0030] In the autonomous vehicle driving decision-making system, scientific obstacle classification is the foundation for accurate decision-making. A sound obstacle classification framework enables the autonomous vehicle to handle different types of objects in a targeted manner, thereby improving decision-making efficiency and safety. The following details the obstacle classification system employed in the solution provided by this disclosure and its role.

[0031] In the disclosed embodiment, first, obstacles are divided into non-interactive obstacles and key obstacles. Non-interactive obstacles refer to objects that have no effect on the driving path of the autonomous vehicle, such as trees in the distance of the road, fixed road signs that do not affect the lane, etc. Such obstacles do not need to be focused on, which can reduce computing resource consumption. Key obstacles are objects that may interact with the autonomous vehicle and are the core processing targets in the decision-making process. For example, a method for identifying key obstacles is provided, such as Figure 2As shown, based on the spatial geometric relationship between the obstacle and the trajectory of the ego vehicle, obstacles can be classified into overlapping obstacles that interact with the driverless vehicle and non-overlapping obstacles that have potential interaction with the driverless vehicle (such as Figure 2 in a)); Overlapping obstacles can be further classified into point-overlapping obstacles (such as Figure 2 in b)) and line-overlapping obstacles (such as Figure 2 in c)) according to the geometric characteristics of the overlapping area. The above three types of obstacles are all key obstacles that may interact with the driverless vehicle.

[0032] Key obstacles are further subdivided into key static obstacles, abnormal obstacles, and key dynamic obstacles. Key static obstacles include constant static obstacles and active static obstacles. Constant static obstacles refer to objects with fixed positions and stable states, such as road isolation piers and vehicles waiting statically in front of a stop line; Active static obstacles, although in a static state, have the potential for dynamic changes. For example, a vehicle parked temporarily by the roadside with the door not closed may have a passenger suddenly getting out later, and the driverless vehicle needs to be vigilant about it and reserve space for response.

[0033] Abnormal obstacles are objects that do not conform to conventional motion laws or have unpredictable behaviors, such as animals suddenly entering the road, falling goods, etc. Such obstacles require the driverless vehicle to react quickly and re-plan the path. Key dynamic obstacles are objects with obvious moving characteristics and may interact with the driverless vehicle, such as motor vehicles changing lanes and pedestrians crossing the road. The driverless vehicle needs to track their dynamics in real time and adjust its own driving strategy to ensure safe interaction.

[0034] In the embodiments of the present disclosure, various types of objects that may be encountered during the driving of the driverless vehicle are systematically classified, and through differential processing logics, accurate responses to obstacles with different risk levels are achieved. Compared with simple obstacle discrimination methods, the classification provided by the present disclosure can help the driverless vehicle quickly identify potential threats in complex environments, reasonably allocate computing resources, effectively reduce the collision risk, and significantly improve the decision-making accuracy and driving safety of the driverless vehicle in complex road conditions.

[0035] In another embodiment provided by the present disclosure, the above step S103, generating a sampling path in a preset area range in the driving direction of the driverless vehicle and performing a collision detection on the sampling path based on key obstacles of a first preset type, and screening out the first sampling path, can be implemented as: Step 1: Generate sampling points in a preset area range in the driving direction of the driverless vehicle, and perform a first collision detection on the sampling points based on key obstacles of a first preset type, and screen out the first sampling points; Step 2: Generate multiple sampling paths based on the first sampling points, and perform a second collision detection on the multiple sampling paths based on key obstacles of a first preset type, and screen out the first sampling path.

[0036] In the decision-making process of an autonomous vehicle, path planning is a crucial step to ensure safe driving. Screening sampling paths based on key obstacles of the first preset type is the core step to optimize the driving route while ensuring safety. To improve the accuracy and efficiency of path screening, this step can be refined into two stages: sampling point screening and secondary path screening.

[0037] In the embodiments of the present disclosure, first in the first stage, sampling points are generated within a preset area range in the driving direction of the autonomous vehicle, and a first collision detection is performed on the sampling points based on key obstacles of the first preset type (such as static obstacles like stationary vehicles and fixed isolation piers). The first sampling points are screened out. This process is similar to marking potential feasible positions on a map, and through the collision detection algorithm, it quickly determines whether each sampling point will collide with key obstacles of the first preset type, excluding the points with collision risks. For example, at the entrance and exit of a parking lot, if there are stationary vehicles occupying part of the passage, this step can quickly exclude the sampling points within the area occupied by the vehicles and retain the points within the safe area as the first sampling points. This step significantly reduces the range of subsequent path generation, reduces the computational amount, and lays a foundation for efficient decision-making.

[0038] Then in the second stage, a second collision detection is performed on multiple sampling paths generated from the first sampling points based on key obstacles of the first preset type, and the first sampling path is screened out. Based on the first sampling points, all the first sampling points are connected to generate sampling paths, such as straight-line driving and turning at different angles. Collision detection is performed again on key obstacles of the first preset type to ensure that there is no collision risk throughout the path. For example, when driving on a narrow street, all possible sampling paths are generated according to the screened first sampling points. After the second collision detection, the safe and reasonable detour paths are retained as the first sampling paths. This secondary detection further optimizes the path plan and improves the reliability of path planning.

[0039] In summary, the method provided by the present disclosure splits the sampling path screening into sampling point screening and secondary path screening, and adopts a hierarchical and progressive processing logic. Compared with a single path screening method, this solution reduces the overall computational complexity by excluding risk points and paths in stages, improves the accuracy and safety of path planning, can help the autonomous vehicle quickly find a safe and feasible driving path in a complex environment, effectively reduces the risk of collision with static or low-speed obstacles, and improves the passing ability and decision-making efficiency of the autonomous vehicle in complex road conditions.

[0040] In another embodiment provided by the present disclosure, the preset area range includes: a longitudinal length and a transverse width; In step 1 above, generating sampling points within a preset area range in the driving direction of the autonomous vehicle can be implemented as: Step (1): Determine the longitudinal sampling interval on the longitudinal length based on the longitudinal remaining length of the predicted trajectory of the driverless vehicle and / or the longitudinal position of the key obstacle; Step (2): Generate longitudinal sampling points on the longitudinal length in the driving direction of the driverless vehicle according to the longitudinal sampling interval; Step (3): Determine the lateral sampling interval on the lateral width corresponding to each longitudinal sampling point based on the lateral position of the key obstacle and the lateral width of the drivable space in the lateral direction of each longitudinal sampling point; Step (4): Generate lateral sampling points in the lateral direction of the position where each longitudinal sampling point is located according to the corresponding lateral sampling interval; Among them, the longitudinal length is determined based on the driving speed of the driverless vehicle; the lateral width is determined based on the lateral drivable boundary of the driverless vehicle, and it satisfies that the closer to the driverless vehicle, the narrower the lateral width.

[0041] During the process of path planning by the driverless vehicle based on key obstacles, the generation of sampling points within the preset area range is the core basic link. The setting of the longitudinal length and lateral width of this area directly affects the rationality of the sampling point distribution and the accuracy of subsequent path planning. Through the systematic sampling point generation process, the driverless vehicle can more accurately capture the feasible space and provide a reliable guarantee for safe driving.

[0042] In the embodiments of the present disclosure, first, determine the longitudinal sampling interval on the longitudinal length based on the longitudinal remaining length of the predicted trajectory of the driverless vehicle and / or the longitudinal position of the key obstacle. If there is no key obstacle at a long distance ahead of the predicted trajectory of the driverless vehicle, a larger longitudinal sampling interval can be set; if approaching a key obstacle, such as a fleet waiting to pass at the intersection ahead, the sampling interval is reduced to more densely capture potential path points. For example, on a highway where the vehicle has a wide field of vision ahead, the longitudinal sampling interval can be set to 5 meters; while when approaching the vehicle in front in a congested section, the interval may be reduced to 1 meter to balance the calculation efficiency and the accuracy of path planning.

[0043] Subsequently, generate longitudinal sampling points within the preset longitudinal length range in the driving direction of the driverless vehicle according to the longitudinal sampling interval to form preliminary longitudinal path nodes, providing a basis for subsequent lateral sampling.

[0044] Then, determine the lateral sampling interval on the lateral width corresponding to each longitudinal sampling point based on the lateral position of the key obstacle and the lateral width of the drivable space in the lateral direction of each longitudinal sampling point. For example, when the driverless vehicle needs to bypass a vehicle parked by the roadside, the lateral drivable space near the vehicle is narrow, and the lateral sampling interval is set small; the space far from the vehicle is wide, and the interval is set large, so as to accurately adapt to the passing conditions at different positions.

[0045] Finally, in the horizontal direction at the position of each longitudinal sampling point, horizontal sampling points are generated according to the corresponding horizontal sampling interval to complete the overall layout of sampling points within the preset area. Among them, the longitudinal length is determined according to the driving speed of the unmanned vehicle. The faster the speed, the longer the preset longitudinal range; the horizontal width is based on the lateral drivable boundary of the unmanned vehicle, and follows the rule that the closer to the unmanned vehicle, the narrower the horizontal width, and thus the smaller the horizontal sampling interval closer to the unmanned vehicle, simulating the space limitation in the real driving scenario.

[0046] In summary, by dynamically adjusting the longitudinal and horizontal sampling intervals, combining multiple factors such as the driving speed of the unmanned vehicle, the position of obstacles, and the drivable boundary, the adaptive distribution of sampling points is achieved. Compared with fixed-interval sampling, it improves the sampling flexibility and environmental adaptability, can capture the feasible path space more accurately, reduce invalid sampling, improve the path planning efficiency and accuracy, and effectively ensure the safe and efficient driving of the unmanned vehicle under complex road conditions.

[0047] In another embodiment provided by the present disclosure, the key obstacles of the first preset type include abnormal obstacles and key static obstacles; In the above step 1, the first collision detection of the sampling points based on the key obstacles of the first preset type can be implemented as: Step a: Determine whether there is an overlap between the sampling point and the key obstacles of the first preset type. If there is an overlap, remove the sampling point to obtain the first sampling point that does not overlap with the key obstacles of the first preset type; Step b: Perform a second collision detection on multiple sampling paths based on the key obstacles of the first preset type, and screen out the first sampling path, including: Step c: Determine whether there is an intersection between the multiple sampling paths and the key obstacles of the first preset type. If there is an intersection, remove the sampling path to obtain the first sampling path that does not intersect with the key obstacles of the first preset type.

[0048] During the path planning process of the unmanned vehicle, collision detection is the core link to ensure driving safety. Conducting collision detection on the key obstacles of the first preset type (including abnormal obstacles and key static obstacles) can effectively eliminate dangerous paths and ensure the safe passage of the unmanned vehicle through complex road conditions. This detection process forms a strict hierarchical screening mechanism through the screening at two levels of sampling points and sampling paths.

[0049] In the embodiment of the present disclosure, first, at the sampling point level, it is necessary to determine whether there is an overlap between the sampling point and the key obstacles of the first preset type. If there is an overlap, it means that the sampling point is in the space occupied by the obstacle, and it will be removed, so as to obtain the first sampling point that does not overlap with the key obstacles of the first preset type. For example Figure 3As shown, the sampling points marked with cross signs overlap with obstacles and should be removed. For example, when there are fallen goods (abnormal obstacles) on the road or parked vehicles (key static obstacles) by the roadside, this step can quickly eliminate the sampling points located at the positions of the goods or vehicles. This operation can initially filter out obviously infeasible path nodes and reduce the computational complexity of subsequent path planning.

[0050] Next, after generating multiple sampling paths based on the selected first sampling points, enter the collision detection stage of the sampling paths. This stage needs to determine whether there is an intersection between the multiple sampling paths and the key obstacles of the first preset type. If there is an intersection (as shown by the dotted line in Figure 3 ), it means that the path will collide with the obstacle during driving, and it will be removed. Finally, the first sampling path that does not intersect with the key obstacles of the first preset type is obtained. For example, in a narrow lane, among the multiple detour paths generated based on the first sampling points, if a certain path is planned to pass through the area where the parked vehicle is located, this path will be excluded, and the path plan that safely avoids the obstacle will be retained.

[0051] The present disclosure adopts a hierarchical collision detection method. First, a preliminary screening is carried out at the sampling point level, and then a deep investigation is carried out on the sampling paths. Compared with a single collision detection method, it realizes the avoidance of abnormal obstacles and key static obstacles, and improves the safety and reliability of path planning. By efficiently filtering dangerous paths, it can help the unmanned vehicle quickly find a safe and feasible driving route in a complex environment and effectively reduce the collision risk.

[0052] In another embodiment provided by the present disclosure, after obtaining the first sampling points in step 1, it further includes: Excluding the first sampling points that exceed the steering ability of the unmanned vehicle from the first sampling points to obtain the first sampling points after re-screening; Furthermore, in step 2, generating multiple sampling paths based on the first sampling points can be implemented as: Generating multiple sampling paths based on the first sampling points after re-screening.

[0053] In the path planning process of the unmanned vehicle, to ensure the practical feasibility of the generated sampling paths, it can be further screened in combination with the performance of the unmanned vehicle itself. Therefore, in the embodiment of the present disclosure, after completing the collision detection based on the key obstacles of the first preset type to obtain the first sampling points, then consider the steering ability of the unmanned vehicle, exclude the sampling points that exceed the steering ability, and generate sampling paths based on the results after re-screening to achieve the matching of path planning and the actual control performance of the vehicle.

[0054] In the embodiments of the present disclosure, after obtaining the first sampling points, these sampling points are screened again to exclude the first sampling points that exceed the turning ability of the driverless vehicle. The turning ability of the driverless vehicle is restricted by factors such as the vehicle's mechanical structure and power system. If the distance between sampling points is too large or the angular deviation exceeds the vehicle's turning limit, even if there is no risk of obstacle collision in space for this path, it cannot be actually executed. For example, in a narrow alley scenario, if the turning angle set for a certain first sampling point exceeds the maximum turning angle of the driverless vehicle and the vehicle cannot complete the corresponding turning action, this sampling point will be excluded. This step incorporates the vehicle's physical performance parameters into the screening conditions to ensure that the remaining sampling points meet the actual controllability feasibility of the driverless vehicle.

[0055] Subsequently, multiple sampling paths are generated based on the first sampling points after the second screening. These sampling points that have passed the turning ability verification serve as the basis for path generation and can ensure that the generated paths can be executed by the driverless vehicle during actual driving. For example, in the scenario of backing out of a parking space in a parking lot, the sampling points screened by the turning ability will guide the generation of a backing path that conforms to the vehicle's turning characteristics, avoiding the planning of sharp turns or large-amplitude turning paths that the vehicle cannot achieve.

[0056] In the embodiments of the present disclosure, the constraint conditions of the turning ability of the driverless vehicle are incorporated into the process of screening the sampling points. This not only ensures that the path is safe and collision-free but also guarantees the executability of the path from the perspective of the vehicle's actual control, enhancing the practicality and reliability of the driverless vehicle path planning. By accurately matching the environmental feasibility and vehicle performance, it effectively reduces decision-making errors caused by the mismatch between path planning and vehicle capabilities, and enhances the passing ability and operation stability of the driverless vehicle in complex road conditions.

[0057] In another embodiment provided by the present disclosure, the second preset type of key obstacles includes key dynamic obstacles; The above step S104, determining the interaction intention between the driverless vehicle and the second preset type of key obstacles according to the first dynamic intention of the second preset type of key obstacles and the second dynamic intention of the driverless vehicle when driving on the first sampling path, can be implemented as: Step 1: Obtain the predicted trajectory of the key dynamic obstacle, and sample multiple accelerations of the key dynamic obstacle when driving on the predicted trajectory to obtain the first dynamic intention; Step 2: For each first sampling path, sample multiple accelerations of the driverless vehicle when driving on this first sampling path to obtain the second dynamic intention; Step 3: Combine the first dynamic intention and the second dynamic intention with the first sampling path to determine the interaction intention of the driverless vehicle to avoid interacting with the key dynamic obstacle during the driving process on the first sampling path.

[0058] After the path of the driverless vehicle is initially screened based on static and abnormal obstacles, in the face of the dynamic interaction scenario of key dynamic obstacles, it is necessary to further accurately determine the driving strategy. For the key obstacles of the second preset type (i.e., key dynamic obstacles), by sampling and comprehensively analyzing their dynamic intentions with the driverless vehicle, potential interaction risks can be effectively predicted, reasonable interaction intentions can be formulated, and the safe passage of the driverless vehicle in the dynamic traffic environment can be ensured.

[0059] In the embodiments of the present disclosure, first, obtain the predicted trajectory of the key dynamic obstacle, and as Figure 4 and Figure 5 shown, sample various accelerations of the key dynamic obstacle during driving on the predicted trajectory to obtain the first dynamic intention. For example, at an intersection, a motor vehicle approaching horizontally is used as a key dynamic obstacle. The driving trajectory of the motor vehicle in the next period of time is obtained through sensor data and a motion prediction model, and various acceleration changes such as possible acceleration, deceleration, and constant speed are sampled. This step can comprehensively capture the potential motion changes of the key dynamic obstacle, provide rich data support for subsequent interaction analysis, and avoid misjudging risks due to only considering a single motion state.

[0060] Next, for each first sampled path, sample various accelerations of the driverless vehicle during driving on this path to obtain the second dynamic intention. Taking the driverless vehicle planning to go straight through the intersection as an example, sample different acceleration states such as possible acceleration, deceleration, and maintaining a constant speed on this first sampled path, so as to clarify the possibility of the dynamic behavior of the driverless vehicle itself on this path and ensure that the analysis of the driving intention of the driverless vehicle meets the requirements of the actual driving scenario.

[0061] Finally, combine the first dynamic intention and the second dynamic intention with the first sampled path to determine the interaction intention of the driverless vehicle to avoid collision with the key dynamic obstacle during driving on the first sampled path. By analyzing the spatio-temporal relationship between the key dynamic obstacle and the driverless vehicle in their respective different acceleration states, select a solution that will not cause trajectory intersection or collision risk. For example, in the above intersection scenario, if it is found that the driverless vehicle maintaining a constant speed and going straight will have a collision risk with the motor vehicle approaching at an accelerated speed, then through combined analysis, determine a reasonable driving strategy for the driverless vehicle to decelerate and give way or accelerate in advance to avoid interaction.

[0062] In the embodiments of the present disclosure, through multi-dimensional sampling and combined analysis of the dynamic intentions of the driverless vehicle and the key dynamic obstacle, incorporating dynamic behavior changes into the interaction decision-making system, potential risks in complex traffic scenarios can be predicted more accurately, and practical interaction intentions can be formulated. It effectively improves the decision-making scientificity and safety of the driverless vehicle in the dynamic traffic environment, reduces collision accidents caused by the uncertainty of the behavior of dynamic obstacles, and enhances the traffic efficiency and reliability of the driverless vehicle in complex road conditions.

[0063] In another embodiment provided by the present disclosure, the above step S105 of evaluating the interaction intention and generating a driving strategy for the driverless vehicle according to the evaluation result can be implemented as follows: Step 1: Evaluate the preset indicators of each interaction intention; among them, the preset indicators include at least one of the following: safety, anthropomorphism, and passability; Step 2: Evaluate and sort the interaction intentions according to the evaluation results of each interaction intention; Step 3: Classify and code the sorted interaction intentions based on multi-intention requirements, and eliminate duplicate intentions with lower rankings for each classification code; Step 4: Generate a driving strategy for the driverless vehicle based on the remaining interaction intentions after eliminating duplicate intentions and the evaluation and sorting results of the remaining interaction intentions.

[0064] After determining the interaction intentions between the driverless vehicle and key dynamic obstacles, these intentions need to be converted into executable driving strategies. To this end, by systematically evaluating and processing the interaction intentions, the advantages and disadvantages of the intentions are measured from multiple dimensions, screened and optimized, and finally a driving strategy that meets the actual needs is generated to ensure the safe and efficient driving of the driverless vehicle in complex traffic scenarios.

[0065] In the embodiment of the present disclosure, first, the preset indicators of each interaction intention are evaluated. The preset indicators cover aspects such as safety, anthropomorphism, and passability. The safety indicator evaluates whether the interaction intention will cause the driverless vehicle to collide with an obstacle. For example, for the intention of the driverless vehicle to decelerate and avoid a pedestrian, it is necessary to judge whether the driving trajectory after deceleration is safe; the anthropomorphism indicator considers whether the intention conforms to human driving habits, such as avoiding abrupt behaviors such as sudden acceleration and sudden braking; the passability indicator evaluates whether this intention can enable the driverless vehicle to smoothly pass through the current section, such as the passing efficiency in a narrow passage. This step provides a quantitative basis for subsequent screening.

[0066] Next, according to the evaluation results of each interaction intention, the interaction intentions are evaluated and sorted. For example, in the scenario of avoiding pedestrians at an intersection, the interaction intentions with high safety, in line with human driving habits, and good passability are ranked at the front, while the intentions with collision risks or being too abrupt are ranked at the back, thus forming a priority order for subsequent screening decisions.

[0067] Subsequently, based on multi-intention requirements, the sorted interaction intentions are classified and coded, and duplicate intentions with lower rankings are eliminated for each classification code. For example, intentions such as "decelerate and give way" and "pass at a constant speed" are classified and coded. If there are multiple similar intentions in a certain classification code, only the one with a higher ranking is retained, and the duplicate and less effective intentions are removed to reduce redundant decision-making information and improve decision-making efficiency.

[0068] Finally, a driving strategy for the driverless vehicle is generated based on the remaining interaction intentions after removing duplicate intentions and the evaluation and ranking results of the remaining interaction intentions. Interaction intentions with higher rankings are selected to generate a driving strategy with execution priorities to guide the actual driving of the driverless vehicle. When selecting, one optimal interaction intention can be selected, or a preset number of interaction intentions with relatively high rankings can be selected. There is no limitation here.

[0069] In summary, the method provided by the embodiments of the present disclosure screens out relatively excellent driving decisions for driverless vehicles through multi-dimensional evaluation, ranking and screening, and classification optimization. Compared with single-index evaluation, it takes into account multiple aspects such as safety, anthropomorphism, and passability, balances safety and traffic efficiency, and reduces duplicate and ineffective decisions at the same time. It effectively improves the rationality and reliability of the driverless vehicle's decision-making, enables it to make driving decisions closer to human driving habits, safer and more efficient in complex traffic environments, and enhances the practicality and user acceptance of the driverless vehicle.

[0070] In another embodiment provided by the present disclosure, before the above step S104, according to the first dynamic intention of the key obstacles of the second preset type and the second dynamic intention of the driverless vehicle, to determine the interaction intention between the driverless vehicle and the key obstacles of the second preset type when driving on the first sampling path, it further includes: Evaluating and ranking the first sampling paths based on at least one of the path shape change variance, the distance from the reference line, the distance from the active static obstacles, and the end point distance of the first sampling path, to obtain the first sampling paths ranked in the top preset number; Among them, the smaller the path shape change variance, the higher the corresponding evaluation value; the smaller the distance from the reference line, the higher the corresponding evaluation value; the larger the distance from the active static obstacles, the higher the corresponding evaluation value; the smaller the end point distance, the higher the corresponding evaluation value; Furthermore, the above step S104, according to the first dynamic intention of the key obstacles of the second preset type and the second dynamic intention of the driverless vehicle when driving on the first sampling path, to determine the interaction intention between the driverless vehicle and the key obstacles of the second preset type, can also be implemented as: According to the first dynamic intention of the key obstacles of the second preset type and the second dynamic intention of the driverless vehicle when driving on the first sampling paths ranked in the top preset number, to determine the interaction intention between the driverless vehicle and the key obstacles of the second preset type.

[0071] After the driverless vehicle completes the preliminary path screening to obtain the first sampling paths, since there are differences in terms of safety, comfort, and feasibility among these paths, directly using them for interaction intention analysis with key dynamic obstacles may lead to low decision-making efficiency or irrationality. Therefore, it is necessary to further quantitatively evaluate and screen the first sampling paths to optimize the subsequent interaction decision-making process with key dynamic obstacles.

[0072] In the embodiments of the present disclosure, first, the first sampling paths are evaluated and sorted based on at least one of the variance of the path shape change of the first sampling path, the distance from the reference line, the distance from the active static obstacle, and the end point distance. The variance of the path shape change reflects the smoothness of the path. The smaller the variance, the smoother the path, the higher the driving comfort of the vehicle, and the higher the corresponding evaluation value. For example, on a straight road, a smooth straight driving path has a smaller variance of shape change and a higher evaluation value compared to a path with frequent steering; the distance from the reference line measures the degree to which the path deviates from the ideal driving route. The smaller the distance, the more in line with the expected driving direction, and the higher the evaluation value; the greater the distance from the active static obstacle, the higher the driving safety, and the corresponding evaluation value is improved. For example, when there are temporarily parked vehicles by the roadside, a path far from the vehicle has better safety and a higher evaluation value; the end point distance refers to the distance between the end point of the path and the target point. The smaller this distance, the closer the path is to the expected target, and the higher the evaluation value.

[0073] For the calculation of the evaluation value, taking the case where all the above four factors are involved in the evaluation as an example, accurate evaluation can be carried out by using the method of calculating weights during implementation. For example, the following formula is used for calculation.

[0074]

[0075] Among them, , , and represent the weight coefficients, represents the variance of the path shape change, represents the distance from the reference line, represents the end point distance, represents the distance from the active static obstacle.

[0076] Through the above evaluation criteria, all the first sampling paths are comprehensively scored and sorted according to the evaluation values from high to low, and the first sampling paths ranked in the top preset number are obtained. For example, in a complex urban street scenario, after screening, the first 5 first sampling paths with the highest evaluation values are retained for the next stage of decision-making.

[0077] Subsequently, according to the first dynamic intention of the key obstacles of the second preset type and the second dynamic intention of the driverless vehicle when driving on the first sampling paths ranked in the top preset number, the interaction intention between the driverless vehicle and the key obstacles of the second preset type is determined. This step aims to generate the interaction intention between the driverless vehicle and the key dynamic obstacles, and through the above implementation method, the interaction decision between the driverless vehicle and the key dynamic obstacles is based on a safer and more reasonable path, reducing unnecessary calculations and decision-making errors.

[0078] In summary, the method provided by the present disclosure screens and optimizes the first sampling path by introducing multi-dimensional path evaluation indicators, which can effectively reduce the computational burden caused by redundant paths and improve the decision-making efficiency. At the same time, by screening out better paths, the interaction decision between the driverless vehicle and key dynamic obstacles becomes more scientific and reasonable, enhancing the driving safety and comfort of the driverless vehicle in a complex traffic environment.

[0079] In another embodiment provided by the present disclosure, the above step S102, when a key obstacle is detected, identifying the type of the key obstacle, may be implemented as: When a key obstacle is detected, identifying the type of the obstacle according to the reference line projection of the key obstacle; Furthermore, the above method further includes: When the obstacle is an abnormal obstacle, expanding the obstacle to form an abnormal driving area as the obstacle corresponding area; When the obstacle is an obstacle that normally interacts with the driverless vehicle, obtaining a specified area near the proximal end of the predicted trajectory of the obstacle as the obstacle corresponding area; Constructing an SL map based on the obstacle corresponding areas of different types of obstacles.

[0080] In the decision-making process of the driverless vehicle, accurately identifying the type of key obstacles is the basis for subsequent targeted processing. Through the identification method based on reference line projection, different types of key obstacles can be efficiently distinguished, and corresponding obstacle corresponding areas can be constructed according to the type differences, and finally an SL map is formed, providing an intuitive basis for the driverless vehicle to plan a safe path. Among them, the SL map is a map constructed by SLAM technology in the SLAM (Simultaneous Localization and Mapping) scenario. It is a digital representation of the surrounding environment by the driverless vehicle. The driverless vehicle uses multiple sensors such as lidar and cameras to collect environmental information, processes these data through the SLAM algorithm, and while determining the vehicle's own position in real time, constructs a map containing information such as roads, buildings, and obstacles. This map can provide a basis for functions such as path planning and obstacle avoidance of the vehicle, helping the vehicle to achieve autonomous navigation in a complex environment.

[0081] In the embodiments of the present disclosure, such as Figure 6As shown in the figure, after detecting a key obstacle, its type is identified based on the projection of the reference line of the key obstacle. The reference line can be set as the center line of the predetermined driving path of the driverless vehicle or the center line of the road. By analyzing the position, shape, and variation law of the obstacle projection on the reference line, its type is judged. For example, if the obstacle projection is near the reference line and has an irregular shape, combined with the sensor data, it can be identified as an abnormal obstacle that suddenly intrudes into the road; if the projection position is stable and conforms to the conventional motion pattern of a vehicle or a pedestrian, it may be an obstacle that normally interacts with the driverless vehicle. This projection-based identification method uses geometric relationships to quickly locate features and reduces the complexity of identification calculations.

[0082] Optionally, after identifying and outputting the labels of each obstacle, when the identified obstacle is an abnormal obstacle, it is inflated to form an abnormal driving area as the corresponding area of the obstacle. For example, when there is suddenly a dropped cargo on the road, the actual occupied space of the cargo is expanded outward according to certain rules to generate a larger abnormal driving area, avoiding collisions caused by changes in the position of the cargo or positioning errors of the driverless vehicle, and reserving sufficient safe detour space for the driverless vehicle.

[0083] Optionally, if the obstacle is an obstacle that normally interacts with the driverless vehicle, a specified area near the proximal end of the predicted trajectory of the obstacle is obtained as the corresponding area of the obstacle. Taking a vehicle traveling in the same direction as an example, an area at a certain distance in front of the predicted trajectory within the next three seconds is taken as the corresponding area, focusing on the spatial range where interactions may occur in the short term, reducing unnecessary spatial calculations, and focusing on the core area of interaction decision-making.

[0084] Finally, an SL map is constructed based on the corresponding areas of different types of obstacles. The corresponding areas of various obstacles are integrated into a unified coordinate system, intuitively presenting the relationship between the obstacles and the driving space of the driverless vehicle, providing a clear spatial reference for subsequent path planning, collision detection, and analysis of interaction intentions, and facilitating the driverless vehicle to quickly judge the feasible path and risk area.

[0085] In summary, the method provided by the present disclosure combines reference line projection recognition with differential area construction. Compared with a single obstacle recognition method, it not only improves the type recognition efficiency, but also enhances the safety and flexibility of decision-making through targeted area processing. By constructing an SL map, the visualization integration of multi-type obstacle information is realized, providing strong support for the efficient decision-making of the driverless vehicle in a complex environment and reducing the accident risk caused by misjudgment or improper handling of obstacles.

[0086] Figure 7 Another embodiment of the decision-making process of the driverless vehicle is also provided. As Figure 7 shown, it includes: First, start the obstacle decision-making process. Distinguish the types of obstacles according to different interaction states (no interaction, normal interaction dynamic, normal interaction static, abnormal). For normal interaction dynamic obstacles, first take the area in the nearly three seconds of the predicted trajectory; for normal interaction static and abnormal obstacles, respectively, through operations such as dilation to cover the driving area, jointly construct the SL map. Subsequently, successively carry out sampling at specific orientation positions of key obstacles and sampling of lateral sampling points based on the driving intention. After collision detection of the sampling points, filter out the valid sampling points, generate the sampling trajectory and construct the corresponding obstacle situation, then perform deduplication of intention coding and acceleration sampling decision for regional dynamic obstacles, and finally conduct comprehensive evaluation of the sampling trajectory. The process ends. Thus, the detection, sampling, and trajectory evaluation of various obstacles are realized, ensuring driving safety and decision-making rationality.

[0087] Based on the same general inventive concept, an embodiment of the present disclosure also provides a decision-making device for an autonomous vehicle. As Figure 8 shown, the device may include: An obstacle detection module 801, configured to detect obstacles in the driving direction of the autonomous vehicle; An obstacle recognition module 802, configured to recognize the type of a key obstacle when the key obstacle is detected; wherein, the key obstacle includes an obstacle that may interact with the autonomous vehicle; A collision test module 803, configured to generate a sampling path within a preset area range in the driving direction of the autonomous vehicle, and perform collision detection on the sampling path based on a first preset type of key obstacle to filter out the first sampling path; An interaction intention module 804, configured to determine the interaction intention between the autonomous vehicle and a second preset type of key obstacle according to the first dynamic intention of the second preset type of key obstacle and the second dynamic intention of the autonomous vehicle when driving on the first sampling path; A strategy generation module 805, configured to evaluate the interaction intention and generate a driving strategy for the autonomous vehicle according to the evaluation result.

[0088] In another embodiment provided by the present disclosure, the obstacles include: non-interacting obstacles and key obstacles; the key obstacles include: key static obstacles, abnormal obstacles, and key dynamic obstacles; the key static obstacles include: constant static obstacles and active static obstacles.

[0089] In another embodiment provided by the present disclosure, the collision test module 803 is configured to generate sampling points within a preset area range in the driving direction of the autonomous vehicle, and perform a first collision detection on the sampling points based on a first preset type of key obstacle to filter out the first sampling points; generate multiple sampling paths based on the first sampling points, and perform a second collision detection on the multiple sampling paths based on the first preset type of key obstacle to filter out the first sampling path.

[0090] In another embodiment provided by the present disclosure, the preset area range includes: a longitudinal length and a transverse width. The collision test module 803 is configured to determine a longitudinal sampling interval on the longitudinal length based on the longitudinal remaining length of the predicted trajectory of the driverless vehicle and / or the longitudinal position of the key obstacle; generate longitudinal sampling points on the longitudinal length in the driving direction of the driverless vehicle according to the longitudinal sampling interval; determine a transverse sampling interval on the transverse width corresponding to each longitudinal sampling point based on the transverse position of the key obstacle and the transverse width of the drivable space in the transverse direction of each longitudinal sampling point; and generate transverse sampling points in the transverse direction of the position where each longitudinal sampling point is located according to the corresponding transverse sampling interval. The longitudinal length is determined based on the driving speed of the driverless vehicle, and the transverse width is determined based on the transverse drivable boundary of the driverless vehicle, and satisfies that the closer to the driverless vehicle, the narrower the transverse width.

[0091] In another embodiment provided by the present disclosure, the key obstacles of the first preset type include abnormal obstacles and key static obstacles. The collision test module 803 is configured to determine whether there is an overlap between the sampling points and the key obstacles of the first preset type. If there is an overlap, the sampling point is removed to obtain first sampling points that do not overlap with the key obstacles of the first preset type; determine whether there is an intersection between multiple sampling paths and the key obstacles of the first preset type. If there is an intersection, the sampling path is removed to obtain first sampling paths that do not intersect with the key obstacles of the first preset type.

[0092] In another embodiment provided by the present disclosure, the collision test module 803 is further configured to exclude first sampling points that exceed the steering ability of the driverless vehicle from the first sampling points to obtain the first sampling points after re-screening; and generate multiple sampling paths based on the first sampling points after re-screening.

[0093] In another embodiment provided by the present disclosure, the key obstacles of the second preset type include key dynamic obstacles. The interaction intention module 804 is configured to obtain the predicted trajectory of the key dynamic obstacle, sample various accelerations of the key dynamic obstacle during the predicted trajectory driving to obtain a first dynamic intention; for each first sampling path, sample various accelerations of the driverless vehicle during the driving of the first sampling path to obtain a second dynamic intention; and combine the first dynamic intention and the second dynamic intention with the first sampling path to determine an interaction intention for the driverless vehicle to avoid interacting with the key dynamic obstacle during the driving of the first sampling path.

[0094] In another embodiment provided by the present disclosure, the policy generation module 805 is configured to evaluate the preset metrics for each interaction intention; wherein the preset metrics include at least one of the following: security, anthropomorphism, and passability; evaluate and rank the interaction intentions according to the evaluation results of each interaction intention; classify and code the ranked interaction intentions based on the multi-intention requirements, and eliminate the duplicate intentions with lower rankings for each classification code; generate a driving policy for the driverless vehicle based on the remaining interaction intentions after eliminating the duplicate intentions and the evaluation and ranking results of the remaining interaction intentions.

[0095] In another embodiment provided by the present disclosure, the interaction intention module 804 is further configured to evaluate and rank the first sampling path based on at least one of the path form change variance, the distance from the reference line, the distance from the active static obstacle, and the end point distance of the first sampling path, and obtain the first sampling paths with the top preset number of rankings; wherein, the smaller the path form change variance, the higher the corresponding evaluation value; the smaller the distance from the reference line, the higher the corresponding evaluation value; the larger the distance from the active static obstacle, the higher the corresponding evaluation value; the smaller the end point distance, the higher the corresponding evaluation value; determine the interaction intention between the driverless vehicle and the key obstacles of the second preset type according to the first dynamic intention of the key obstacles of the second preset type and the second dynamic intention of the driverless vehicle when driving on the first sampling paths with the top preset number of rankings.

[0096] In another embodiment provided by the present disclosure, the obstacle recognition module 802 is configured to identify the type of the obstacle according to the reference line projection of the key obstacle when the key obstacle is detected. The above device further includes: an SL map construction module, configured to inflate the obstacle to form an abnormal driving area as the obstacle corresponding area when the obstacle is an abnormal obstacle; obtain the proximal specified area of the predicted trajectory of the obstacle as the obstacle corresponding area when the obstacle is an obstacle that interacts with the driverless vehicle normally; construct an SL map based on the obstacle corresponding areas of different types of obstacles.

[0097] According to the embodiments of the present application, a vehicle mixed traffic control system is further provided, and the system may include the decision-making device of any of the above driverless vehicles. The specific implementation manners of the device may refer to the explanations in the above embodiments and will not be elaborated herein.

[0098] Through the description of the above embodiments, those skilled in the art can clearly understand that the embodiments of the present disclosure can be implemented by hardware or by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solutions of the embodiments of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present disclosure.

[0099] Those skilled in the art can understand that the drawings are only schematic diagrams of a preferred embodiment, and the modules or processes in the drawings are not necessarily essential for implementing the present disclosure.

[0100] Those skilled in the art can understand that the modules in the device in the embodiment can be distributed in the device in the embodiment according to the description of the embodiment, or can be correspondingly changed and located in one or more devices different from the present embodiment. The modules of the above embodiments can be combined into one module, or further split into multiple sub-modules.

[0101] The serial numbers of the above embodiments of the present disclosure are only for description and do not represent the advantages and disadvantages of the embodiments.

[0102] Obviously, those skilled in the art can make various changes and modifications to the present disclosure without departing from the spirit and scope of the present disclosure. Thus, if these modifications and variations of the present disclosure fall within the scope of the claims of the present disclosure and their equivalent technologies, the present disclosure also intends to include these changes and modifications.

Claims

1. A decision-making method for an unmanned vehicle, characterized in that, Including: Detecting obstacles in the driving direction of the driverless vehicle; When a key obstacle is detected, identifying the type of the key obstacle; wherein, the key obstacle includes an obstacle that may interact with the driverless vehicle; Generating a sampling path within a preset area range in the driving direction of the driverless vehicle, and performing collision detection on the sampling path based on a first preset type of key obstacle to filter out a first sampling path; Determining the interaction intention between the driverless vehicle and a second preset type of key obstacle according to the first dynamic intention of the second preset type of key obstacle and the second dynamic intention of the driverless vehicle when driving on the first sampling path; Evaluating the interaction intention and generating a driving strategy for the driverless vehicle according to the evaluation result.

2. The method according to claim 1, wherein The obstacles include: non-interactive obstacles and key obstacles; the key obstacles include: key static obstacles, abnormal obstacles and key dynamic obstacles; the key static obstacles include: normal static obstacles and active static obstacles.

3. The method according to claim 1 or 2, characterized in that, The generating a sampling path within a preset area range in the driving direction of the driverless vehicle, and performing collision detection on the sampling path based on a first preset type of key obstacle to filter out a first sampling path includes: Generating sampling points within a preset area range in the driving direction of the driverless vehicle, and performing first collision detection on the sampling points based on a first preset type of key obstacle to filter out first sampling points; Generating multiple sampling paths based on the first sampling points, and performing second collision detection on the multiple sampling paths based on a first preset type of key obstacle to filter out a first sampling path.

4. The method according to claim 3, wherein The preset area range includes: a longitudinal length and a lateral width; Generating sampling points within a preset area range in the driving direction of the driverless vehicle includes: Determining a longitudinal sampling interval on the longitudinal length based on the longitudinal remaining length of the predicted trajectory of the driverless vehicle and / or the longitudinal position of the key obstacle; Generating longitudinal sampling points on the longitudinal length in the driving direction of the driverless vehicle according to the longitudinal sampling interval; Determining a lateral sampling interval on the lateral width corresponding to each longitudinal sampling point based on the lateral position of the key obstacle and the lateral width of the drivable space in the lateral direction of each longitudinal sampling point; Generating lateral sampling points in the lateral direction of the position where each longitudinal sampling point is located according to the corresponding lateral sampling interval; Wherein, the longitudinal length is determined based on the driving speed of the driverless vehicle; the lateral width is determined based on the lateral drivable boundary of the driverless vehicle, and satisfies: the closer to the driverless vehicle, the narrower the lateral width.

5. The method according to claim 3, characterized in that The first preset type of key obstacle includes abnormal obstacles and key static obstacles; The performing first collision detection on the sampling points based on a first preset type of key obstacle includes: Determining whether there is an overlap between the sampling point and the first preset type of key obstacle, if there is an overlap, removing the sampling point to obtain a first sampling point that does not overlap with the first preset type of key obstacle; Performing a second collision detection on the multiple sampling paths based on the key obstacles of the first preset type, and screening out the first sampling path, including: Determining whether there is an intersection between the multiple sampling paths and the key obstacles of the first preset type. If there is an intersection, removing the sampling path to obtain the first sampling path that has no intersection with the key obstacles of the first preset type.

6. The method according to claim 3, wherein After obtaining the first sampling points, it further includes: Excluding the first sampling points that exceed the steering ability of the driverless vehicle from the first sampling points to obtain the first sampling points after re-screening; Generating multiple sampling paths based on the first sampling points, including: Generating multiple sampling paths based on the first sampling points after re-screening.

7. The method according to claim 1, wherein The key obstacles of the second preset type include key dynamic obstacles; Determining the interaction intention between the driverless vehicle and the key obstacles of the second preset type according to the first dynamic intention of the key obstacles of the second preset type and the second dynamic intention of the driverless vehicle when driving on the first sampling path, including: Obtaining the predicted trajectory of the key dynamic obstacle, and sampling various accelerations of the key dynamic obstacle when driving on the predicted trajectory to obtain the first dynamic intention; For each first sampling path, sampling various accelerations of the driverless vehicle when driving on the first sampling path to obtain the second dynamic intention; Combining the first dynamic intention and the second dynamic intention with the first sampling path to determine the interaction intention of the driverless vehicle to avoid interacting with the key dynamic obstacle during driving on the first sampling path.

8. The method according to claim 1, wherein Evaluating the interaction intention and generating the driving strategy of the driverless vehicle according to the evaluation result, including: Evaluating the preset indexes of each interaction intention; wherein, the preset indexes include at least one of the following: safety, anthropomorphism, and passability; Evaluating and sorting the interaction intentions according to the evaluation results of each interaction intention; Classifying and coding the sorted interaction intentions based on multi-intention requirements, and removing the duplicate intentions with lower rankings for each classification code; Generating the driving strategy of the driverless vehicle based on the remaining interaction intentions after removing duplicate intentions and the evaluation and sorting results of the remaining interaction intentions.

9. The method according to claim 1, wherein Before determining the interaction intention between the driverless vehicle and the key obstacles of the second preset type according to the first dynamic intention of the key obstacles of the second preset type and the second dynamic intention of the driverless vehicle, it further includes: Evaluating and sorting the first sampling paths based on at least one of the path form change variance, distance from the reference line, distance from the active static obstacle, and end point distance of the first sampling path to obtain the first sampling paths with the top preset number of rankings; Wherein, the smaller the path form change variance, the higher the corresponding evaluation value; the smaller the distance from the reference line, the higher the corresponding evaluation value; the larger the distance from the active static obstacle, the higher the corresponding evaluation value; the smaller the end point distance, the higher the corresponding evaluation value; Determining the interaction intention between the driverless vehicle and the key obstacle of the second preset type based on the first dynamic intention of the key obstacle of the second preset type and the second dynamic intention of the driverless vehicle when driving on the first sampling path, includes: Determining the interaction intention between the driverless vehicle and the key obstacle of the second preset type based on the first dynamic intention of the key obstacle of the second preset type and the second dynamic intention of the driverless vehicle when driving on the first sampling paths with the preset number ranked in the front.

10. The method according to claim 1, wherein The type recognition of the key obstacle in the case of detecting the key obstacle includes: In the case of detecting the key obstacle, identifying the type of the obstacle according to the reference line projection of the key obstacle; The method further includes: In the case that the obstacle is an abnormal obstacle, expanding the obstacle to form an abnormal driving area as the obstacle corresponding area; In the case that the obstacle is an obstacle that normally interacts with the driverless vehicle, obtaining the proximal designated area of the predicted trajectory of the obstacle as the obstacle corresponding area; Constructing an SL map based on the obstacle corresponding areas of different types of obstacles.

11. A decision-making device for an unmanned vehicle, characterized in that, Including: An obstacle detection module for detecting obstacles in the driving direction of the driverless vehicle; An obstacle recognition module for identifying the type of the key obstacle in the case of detecting the key obstacle; wherein, the key obstacle includes an obstacle that may interact with the driverless vehicle; A collision test module for generating sampling paths in a preset area range in the driving direction of the driverless vehicle and performing collision detection on the sampling paths based on the key obstacle of the first preset type to screen out the first sampling paths; An interaction intention module for determining the interaction intention between the driverless vehicle and the key obstacle of the second preset type based on the first dynamic intention of the key obstacle of the second preset type and the second dynamic intention of the driverless vehicle when driving on the first sampling path; A strategy generation module for evaluating the interaction intention and generating a driving strategy for the driverless vehicle according to the evaluation result.

12. A vehicle mixed traffic control system, characterized in that, Including a decision-making device for a driverless vehicle as described in claim 11.

Images