Automatic driving path planning method and device based on boundary expansion and fusion planning

Abstract

The application provides an automatic driving path planning method and device based on boundary expansion and fusion planning, which comprises the following steps: S1, constructing a plurality of path boundaries based on a current reference line, the current reference line being a center line of a lane where a vehicle is located; S2, screening obstacles for the plurality of path boundaries based on obstacle lane information and obstacle motion intention provided by a decision module; S3, compressing related boundaries of the plurality of path boundaries by using the screened obstacles in step S2; S4, optimizing and solving the path by using a quadratic programming construction method; S5, transmitting path evaluation results to speed planning, so as to avoid collision with dynamic obstacles by speed planning, and evaluating the smoothness of the path; and S6, outputting path planning results. The application can ensure that path optimization is successful for each frame, improve path planning stability, and increase the probability of finding a globally optimal or approximately optimal trajectory based on a plurality of paths for speed planning.

Description

Technical Field

[0001] This invention belongs to the field of autonomous driving technology, specifically relating to an autonomous driving path planning method and apparatus based on boundary expansion and fusion planning. Background Technology

[0002] Autonomous driving technology has developed rapidly over the past decade, and planning is a crucial component. Currently, the industry's commonly used spatiotemporally decoupled planning scheme decomposes planning into path planning and velocity planning. Path planning is responsible for outputting a smooth, collision-free path. In structured urban roads, optimization-based path planning relies on lane information provided by high-precision maps to construct initial path boundaries in the Frenet coordinate system to ensure that autonomous vehicles stay within their lanes. In addition, path planning requires obstacle information from perception and prediction modules to construct collision constraints to ensure that collisions with other road users are avoided.

[0003] The path planning techniques mentioned above have the following shortcomings:

[0004] First, the lane information provided by high-precision maps is not completely accurate, which leads to unstable path planning when separating or merging lanes. These lanes may be narrower than the autonomous vehicle, resulting in path optimization failure. Alternatively, these zigzag lanes may cause large curvature changes in the optimized path, which is not friendly to control and tracking, causing the vehicle to drive outside the lane boundary.

[0005] Second, the decoupling of path planning and speed planning is not intelligent enough in complex dynamic environments because the safety assessment of the path is too coarse, which may filter out the optimal path. For example, if another car slightly occupies the driver's lane and there are vehicles in the adjacent lane, two vehicles are queuing for a red light on the left front, and there are vehicles constantly moving on the right, the optimal path will be filtered out and other paths will be selected and passed to speed planning, causing the driver to remain stuck in the current position. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention discloses an autonomous driving path planning method and apparatus based on boundary expansion and fusion planning.

[0007] On one hand, this invention discloses an autonomous driving path planning method based on boundary expansion and fusion planning, comprising the following steps:

[0008] S1: Construct multiple path boundaries based on the current reference line, where the current reference line is the center line of the lane where the vehicle is located;

[0009] S2: Based on the obstacle lane information and obstacle movement intentions provided by the decision module, obstacles are filtered for the boundaries of multiple paths;

[0010] S3: Compress the relevant boundaries of multiple path boundaries using the obstacles selected in step S2;

[0011] S4: The path is optimized and solved using a quadratic programming construction method;

[0012] S5: Pass the path evaluation results to velocity planning to avoid collisions with dynamic obstacles and evaluate the smoothness of the path.

[0013] S6: Output the path planning results.

[0014] Based on the above technical solution, the following improvements can be made:

[0015] As a preferred option, the obstacle lane information provided by the decision module includes: leading in front of the vehicle in the reference line lane, following behind the vehicle in the reference line lane, RIGTH_PARAL in the right lane of the reference line lane, and LEFT_PARAL in the left lane of the reference line lane.

[0016] As a preferred option, the obstacle movement intentions provided by the decision module include: QUEUE, PARKING, CUT_IN (other vehicles entering the driver's lane from other lanes), and LANE_KEEP (lane keeping).

[0017] As a preferred embodiment, the obstacle classification screened in step S2 includes:

[0018] NON: Path-independent obstacle;

[0019] HARD: Path boundary related obstacles, used to compress the path boundary;

[0020] SOFT: Indicates obstacles along the path, used to change the path.

[0021] As a preferred embodiment, the obstacle screening method in step S2 includes at least one of the following:

[0022] Vertical filtering: Filters out obstacles beyond the length of the path in front of the vehicle and within a certain distance behind the vehicle;

[0023] Lateral filtering: Based on the obstacle lane information provided by the decision module, only obstacles in three lanes are retained;

[0024] Obstacles are selected based on their intent and speed information.

[0025] As a preferred embodiment, in step S5, if the path is not smooth when evaluating its smoothness, a smoothing method is used to optimize the path.

[0026] As a preferred embodiment, each of the constructed multiple path boundaries includes a left boundary and a right boundary. In S3, the boundary compression of the multiple path boundaries is to compress either the left boundary or the right boundary of each path boundary.

[0027] On the other hand, the present invention also discloses an autonomous driving path planning device based on boundary expansion and fusion planning, comprising the following modules:

[0028] The path boundary construction module is used to construct multiple path boundaries based on the current reference line, which is the centerline of the lane where the vehicle is located.

[0029] The obstacle screening module is used to screen obstacles for multiple path boundaries based on obstacle lane information and obstacle movement intentions provided by the vehicle's decision module.

[0030] The path boundary compression module is used to compress the relevant boundaries of multiple paths by utilizing obstacles filtered by the obstacle filtering module.

[0031] The path optimization solution module is used to optimize and solve paths using a quadratic programming construction method.

[0032] The speed planning and path smoothing modules are used to pass the path evaluation results to speed planning, so as to avoid collisions with dynamic obstacles through speed planning and to evaluate the smoothness of the path.

[0033] The path output module is used to output the path planning results.

[0034] This invention discloses an autonomous driving path planning method and apparatus based on boundary expansion and fusion planning, which has the following beneficial effects:

[0035] First, this invention can solve the problem of path optimization failure or large curvature changes caused by inaccurate lane information in high-precision maps, ensuring that path optimization is successful in every frame and improving the stability of path planning.

[0036] Second, given that a globally optimal solution exists for the trajectory, we find the optimal solution for each path in multiple local spaces, and then perform velocity planning based on multiple paths to increase the probability of finding the globally optimal trajectory or a globally approximate optimal trajectory. Attached Figure Description

[0037] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This is an overall flowchart of the method of the present invention;

[0039] Figure 2 This is a schematic diagram of the multi-path boundary construction of the present invention;

[0040] Figure 3 (a)-(c) are schematic diagrams of the present invention using filtered path-related obstacles to compress the boundary;

[0041] Figure 4 This is a schematic diagram of the overall process of the method of the present invention;

[0042] Figure 5 This is a schematic diagram of obstacle screening according to the present invention. Detailed Implementation

[0043] The preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0045] Using ordinal numbers such as “first,” “second,” “third,” etc. to describe ordinary objects merely indicates different instances of similar objects and is not intended to imply that the objects being described must have a given order in time, space, sequence, or any other way.

[0046] Furthermore, the expression "includes" is an "open-ended" expression, which means only that there is a corresponding component or step, and should not be interpreted as excluding additional components or steps.

[0047] To achieve the objectives of this invention, some embodiments of the autonomous driving path planning method and apparatus based on boundary expansion and fusion planning, such as Figure 1 and Figure 4 As shown, the autonomous driving path planning method includes the following steps:

[0048] S1: Construct multiple path boundaries based on the current reference line, where the current reference line is the center line of the lane where the vehicle is located;

[0049] Each of the constructed multiple path boundaries includes a left boundary and a right boundary, such as Figure 2 As shown, three path boundaries are constructed: path boundary 1, path boundary 2, and path boundary 3. Specifically:

[0050] Path boundary 1: The left boundary is the left boundary of lane 2, and the right boundary is the right boundary of lane 2;

[0051] Path boundary 2: The left boundary is the left boundary of lane 1, and the right boundary is the right boundary of lane 2;

[0052] Path boundary 3: The left boundary is the left boundary of lane 2, and the right boundary is the right boundary of lane 3;

[0053] S2: Based on the obstacle lane information and obstacle movement intentions provided by the decision module, obstacles are filtered for the boundaries of multiple paths;

[0054] S3: Compress the relevant boundaries of multiple path boundaries using the obstacles selected in step S2. This compression involves compressing either the left or right boundary of each path boundary. Figure 3 (a) is the path boundary diagram 1. Figure 3 (b) is the path boundary 2 diagram. Figure 3 (c) is a path boundary diagram (3);

[0055] S4: The path is optimized and solved using a quadratic programming construction method;

[0056] S5: Pass the path evaluation results to velocity planning to avoid collisions with dynamic obstacles and evaluate the smoothness of the path.

[0057] S6: Output the path planning results.

[0058] The decision-making module can be, but is not limited to, an existing module within the SOC.

[0059] Furthermore, in some specific embodiments, in S2, the obstacle lane information provided by the decision module includes: leading in front of the vehicle in the reference line lane, following behind the vehicle in the reference line lane, RIGTH_PARAL in the right lane of the reference line lane, and LEFT_PARAL in the left lane of the reference line lane.

[0060] Furthermore, the obstacle movement intentions provided by the decision module include: QUEUE, PARKING, CUT_IN (other vehicles entering the driver's lane from other lanes), and LANE_KEEP (lane keeping).

[0061] Furthermore, the obstacle classifications screened in step S2 include:

[0062] NON: Path-independent obstacle;

[0063] HARD: Path boundary related obstacles, used to compress the path boundary;

[0064] SOFT: Indicates obstacles along the path, used to change the path.

[0065] Furthermore, the obstacle screening method in step S2 includes at least one of the following, wherein it is assumed that the left boundary of lane 1 and the right boundary of lane 3 are fences:

[0066] Vertical filtering: Filters out obstacles beyond the length of the path in front of the vehicle and within a certain distance behind the vehicle;

[0067] Lateral filtering: Based on the obstacle lane information provided by the decision module, only obstacles in three lanes are retained;

[0068] Obstacles are filtered based on their intended direction and speed information, such as... Figure 5 As shown, obstacles in all three lanes must be considered.

[0069] First, determine whether it is a static or dynamic obstacle;

[0070] If it is a static obstacle, based on the lateral distance, if it occupies the corresponding path boundary, it is a HARD type obstacle; otherwise, it is a NON type obstacle.

[0071] If it is a dynamic obstacle and is not within the path boundary, then according to the obstacle's intent, LANE_KEEP and being close to the path boundary are SOFT type obstacles, and CUT_IN are NON type obstacles.

[0072] Furthermore, in step S4, a quadratic programming method is used to optimize the path. The quadratic programming method can be, but is not limited to, the method mentioned in the literature "Optimal Vehicle Path Planning Using Quadratic Optimization for Baidu Apollo Open Platform".

[0073] like Figure 3 As shown, only path boundary 3 can be optimized to find a way around the obstacles queuing in the adjacent lane.

[0074] Furthermore, in step S5, if the path is not smooth when evaluating its smoothness, a smoothing method is used to optimize the path.

[0075] Path planning cannot accurately determine whether a collision with a dynamic obstacle will occur. How to avoid collisions with dynamic obstacles is the output of velocity planning. Path planning cannot utilize this posterior information. A rough assessment of collision safety may filter out the optimal path, so no collision safety assessment is performed.

[0076] Smoothness assessment is necessary because path optimization in the Frenet coordinate system cannot guarantee smooth path curvature, especially when navigating around obstacles. Therefore, post-validation of path curvature is required. If the path is not smooth, smoothing optimization methods can be used, but are not limited to, those described in the paper "Sharpness Continuous Path optimization and Sparsification for Automated Vehicles".

[0077] Furthermore, this invention also discloses an autonomous driving path planning device based on boundary expansion and fusion planning, comprising the following modules:

[0078] The path boundary construction module is used to construct multiple path boundaries based on the current reference line, which is the centerline of the lane where the vehicle is located.

[0079] The obstacle screening module is used to screen obstacles for multiple path boundaries based on obstacle lane information and obstacle movement intentions provided by the vehicle's decision module.

[0080] The path boundary compression module is used to compress the relevant boundaries of multiple paths by utilizing obstacles filtered by the obstacle filtering module.

[0081] The path optimization solution module is used to optimize and solve paths using a quadratic programming construction method.

[0082] The speed planning and path smoothing modules are used to pass the path evaluation results to speed planning, so as to avoid collisions with dynamic obstacles through speed planning and to evaluate the smoothness of the path.

[0083] The path output module is used to output the path planning results.

[0084] The above modules use the methods in autonomous driving path planning for path planning, which will not be elaborated here.

[0085] This invention discloses an autonomous driving path planning method and apparatus based on boundary expansion and fusion planning, which has the following beneficial effects:

[0086] First, this invention can solve the problem of path optimization failure or large curvature changes caused by inaccurate lane information in high-precision maps, ensuring that path optimization is successful in every frame and improving the stability of path planning.

[0087] Second, given that a globally optimal solution exists for the trajectory, we find the optimal solution for each path in multiple local spaces, and then perform velocity planning based on multiple paths to increase the probability of finding the globally optimal trajectory or a globally approximate optimal trajectory.

[0088] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several modifications and improvements without departing from the inventive concept of the present invention, and these all fall within the protection scope of the present invention.

Claims

1. An autonomous driving path planning method based on boundary expansion and fusion planning, characterized in that, Includes the following steps: S1: Construct multiple path boundaries based on the current reference line, where the current reference line is the center line of the lane where the vehicle is located; Each of the constructed multiple path boundaries includes a left boundary and a right boundary; Specifically, three path boundaries are constructed: the first path boundary, the second path boundary, and the third path boundary. The left boundary of the first path boundary is the left boundary of the second lane, and the right boundary of the first path boundary is the right boundary of the second lane. The left boundary of the second path boundary is the left boundary of the first lane, and the right boundary of the second path boundary is the right boundary of the second lane; The left boundary of the third path boundary is the left boundary of the second lane, and the right boundary of the third path boundary is the right boundary of the third lane. S2: Based on the obstacle lane information and obstacle movement intention provided by the decision module, obstacles are filtered for the boundaries of the multiple paths; The obstacle lane information provided by the decision module includes: leading in front of the vehicle in the reference line lane, following behind the vehicle in the reference line lane, RIGTH_PARAL in the right lane of the reference line lane, and LEFT_PARAL in the left lane of the reference line lane. The obstacle movement intentions provided by the decision module include: QUEUE, PARKING, CUT_IN (other vehicle enters the lane from another lane), and LANE_KEEP (lane keeping). The obstacle categories to be screened include: NON: Path-independent obstacle; HARD: Path boundary related obstacles, used to compress the path boundary; SOFT: Indicates obstacles related to the path, used to change the path. The obstacle screening method in step S2 includes at least one of the following: Vertical filtering: Filters out obstacles beyond the length of the path in front of the vehicle and within a certain distance behind the vehicle; Lateral filtering: Based on the obstacle lane information provided by the decision module, only obstacles in three lanes are retained; Obstacles are filtered based on their intent and speed information, specifically including: If it is a static obstacle, it is classified as a HARD type obstacle if it occupies the corresponding path boundary based on the lateral distance. Otherwise, it is a NON type obstacle; If it is a dynamic obstacle and is not within the path boundary, then according to the obstacle's intent, LANE_KEEP and being close to the path boundary are SOFT type obstacles, and CUT_IN are NON type obstacles; S3: Compress the relevant boundaries of the multiple path boundaries using the obstacles selected in step S2; S4: The path is optimized and solved using a quadratic programming construction method; S5: Pass the path evaluation results to velocity planning to avoid collisions with dynamic obstacles and evaluate the smoothness of the path. S6: Output the path planning results.

2. The autonomous driving path planning method according to claim 1, characterized in that, In step S5, when evaluating the smoothness of the path, if the path is not smooth, a smoothing method is used to optimize the path.

3. The autonomous driving path planning method according to claim 1, characterized in that, Each of the constructed multiple path boundaries includes a left boundary and a right boundary. In S3, the boundary compression of the multiple path boundaries is to compress either the left boundary or the right boundary of each path boundary.

4. An autonomous driving path planning device based on boundary expansion and fusion planning, characterized in that, Includes the following modules: The path boundary construction module is used to construct multiple path boundaries based on the current reference line, which is the centerline of the lane where the vehicle is located. Each of the constructed multiple path boundaries includes a left boundary and a right boundary; Specifically, three path boundaries are constructed: the first path boundary, the second path boundary, and the third path boundary. The left boundary of the first path boundary is the left boundary of the second lane, and the right boundary of the first path boundary is the right boundary of the second lane. The left boundary of the second path boundary is the left boundary of the first lane, and the right boundary of the second path boundary is the right boundary of the second lane; The left boundary of the third path boundary is the left boundary of the second lane, and the right boundary of the third path boundary is the right boundary of the third lane. An obstacle screening module is used to screen obstacles for the boundaries of the multiple paths based on obstacle lane information and obstacle movement intentions provided by the vehicle's decision module. The obstacle lane information provided by the decision module includes: leading in front of the vehicle in the reference line lane, following behind the vehicle in the reference line lane, RIGTH_PARAL in the right lane of the reference line lane, and LEFT_PARAL in the left lane of the reference line lane. The obstacle movement intentions provided by the decision module include: QUEUE, PARKING, CUT_IN (other vehicle enters the lane from another lane), and LANE_KEEP (lane keeping). The obstacle categories to be screened include: NON: Path-independent obstacle; HARD: Path boundary related obstacles, used to compress the path boundary; SOFT: Indicates obstacles related to the path, used to change the path. The obstacle screening module performs at least one of the following methods to screen obstacles: Vertical filtering: Filter out obstacles beyond the length of the path in front of the vehicle and within a certain distance behind the vehicle; Lateral filtering: Based on the obstacle lane information provided by the decision module, only obstacles in three lanes are retained; Obstacles are filtered based on their intent and speed information, specifically including: If it is a static obstacle, it is classified as a HARD type obstacle if it occupies the corresponding path boundary based on the lateral distance. Otherwise, it is a NON type obstacle; If it is a dynamic obstacle and is not within the path boundary, then according to the obstacle's intent, LANE_KEEP and being close to the path boundary are SOFT type obstacles, and CUT_IN are NON type obstacles; The path boundary compression module is used to compress the relevant boundaries of the multiple path boundaries using obstacles filtered by the obstacle filtering module. The path optimization solution module is used to optimize and solve paths using a quadratic programming construction method. The speed planning and path smoothing modules are used to pass the path evaluation results to speed planning, so as to avoid collisions with dynamic obstacles through speed planning and to evaluate the smoothness of the path. The path output module is used to output the path planning results.

Images