Control line-based automatic driving unstructured environment path planning method and system

Abstract

The application discloses a kind of automatic driving unstructured environment path planning method and system based on control line, specifically includes the following steps: initialization starting control line and target control line;From all the sub control lines, the control line with the minimum evaluation function f value is selected as the current control line;With current control line as parent control line, through the action u defined by the current control line succ Sub control line is generated;Target control line is traced back, and the final skeleton tree is generated;Further selection is carried out in the generated skeleton tree to generate the final path.The unstructured path planning method based on control line proposed in the application can ignore the resolution limit of grid map, and better path planning result can be obtained with less resource consumption.The application carries out experiment in simulation scene and real vehicle test, and the result shows that scene robustness, calculation efficiency and path quality are greatly improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of automatic driving path planning, and particularly relates to an automatic driving path planning method and system in unstructured environment based on control line. BACKGROUND

[0002] As one of the important technical components of automatic driving, the path planning problem in unstructured environment has been concerned in recent decades. Due to the complexity and diversity of unstructured environment and the nonholonomic constraint of vehicle, the existing path planning algorithms still cannot meet the requirements in terms of scene robustness and real-time calculation. At the same time, the quality of the planned path still has a certain distance compared with human drivers in terms of driving comfort. Therefore, the path planning problem in unstructured environment still faces great challenges.

[0003] The traditional state tree-based graph search algorithm is a typical method to solve this problem. Each node stores the single pose information of the vehicle and is linked to other nodes. Starting from the current searched node, different motion primitives (line segment, circular arc, spiral, etc.) are used to generate new nodes, and the search process is stopped when the newly generated node meets the target point cutoff condition. Backtracking from the target node can obtain the final path. However, when the number of motion primitives is small and the step size is large, the search tree at the end is sparse, which is difficult to meet the accurate target pose requirement; otherwise, a large number of unreasonable node expansions will be introduced, leading to an exponential growth of the number of searched nodes.

[0004] For the traditional method, the state tree is generated step by step by a bottom-up exploration method. However, through observation of human driver behavior, it can be found that human drivers construct driving strategies in a top-down manner. That is, a rough driving path is first generated, and then the details of each step of vehicle motion are determined. The planning of human drivers is not to move a fixed distance each time, but is more forward-looking. SUMMARY

[0005] In order to solve the problems existing in the prior art, inspired by human drivers, the present application provides a new automatic driving path planning algorithm based on control line. Under the "control" of two adjacent control lines, different motion primitives are spliced to generate a smooth turning path, which makes the two control lines reachable for the automatic driving vehicle. Once the current control line can reach the target control line, the search process is immediately ended, so as to construct a skeleton tree with multiple connected skeletons. After permutation and combination, a path set is obtained from the skeleton tree. The goal of C-Line*planner is to generate a "natural" or "human-like" path, so a more comfortable path can be selected from the path set as the final path.

[0006] In order to achieve the above object, the technical scheme adopted by the present application is: a control line-based automatic driving unstructured environment path planning method, specifically comprising the following steps:

[0007] S1, initialization, initializing a starting control line and a target control line;

[0008] S2, obtaining a current control line, selecting a control line with the minimum evaluation function f value from all sub control lines as the current control line;

[0009] S3, sub control line expansion, taking the current control line as a parent control line, generating a sub control line through a defined action u succ .

[0010] S4, target control line backtracking, target control line backtracking, generating a final skeleton tree;

[0011] S5, final path selection, further selecting a final path in the generated skeleton tree.

[0012] In S1, during initialization, the control line is a set of states with fixed orientation and no collision with obstacles, and the starting node x init (x,y,θ) and the target node x goal (x,y,θ) of the vehicle are used to generate a starting control line and a target control line

[0013] During sub control line expansion, the current control line

[0014] is taken as the parent control line, and the sub control line

[0015] is generated through a defined action u succ . The sub control line detected through reachability is given an evaluation function f value, and the newly generated control line is embedded into the OPEN linked list, and the parent control line that has completed expansion is embedded into the CLOSED linked list.

[0016]

[0017] In S2, the evaluation function f value is given to the sub control line detected through reachability, and the evaluation function f value is defined as:

[0018]

[0019] wherein g is a cost function for the distance from the starting control line, h is a heuristic function for the distance from the target control line, and the calculation of the cost function g is the cost of the current parent control line plus the cost between the current control line and the sub control line:

[0020]

[0021] There are many CC-Turns or C-Turns between two adjacent reachable intersection of current control line and sub control line, the average curve distance is used to represent the cost between current control line and sub control line:

[0022]

[0023] Where R represents the reachability between two adjacent control lines, len i is the length of the i-th CC-Turn or C-Turn;

[0024] The heuristic function h is:

[0025]

[0026] Where represents the shortest straight line distance between the current control line and the target control line.

[0027] In S4, when the latest expanded sub control line meets the reachable intersection condition, the reachability R>0, the search process is ended, and the parent control line is added to the last skeleton tree from the target control line.

[0028] In S5, when the final path is selected, the path set in the generated skeleton tree at the intersection of two control lines is arranged and combined to obtain the path set

[0029]

[0030] A combination with the smallest number of gear shifting points is selected from the path set as the final path.

[0031] During the entire steering motion of the vehicle, circular arcs and spirals are used as steering motion primitives.

[0032] A control line-based automatic driving unstructured environment path planning system, comprising an initialization module, a current control line acquisition module, a sub control line expansion module, a target control line backtracking module and a final path selection module; the initialization module is used for initializing a starting control line and a target control line;

[0033] The current control line acquisition module selects the control line with the minimum evaluation function f value from the sub control line as the current control line;

[0034] ​The child control line expansion module takes the current control line as a parent control line, and generates a child control line through a defined action u succ generating a child control line;

[0035] The target control line backtracking module is used for backtracking the target control line to generate a final skeleton tree.

[0036] The final path selection module further selects in the generated skeleton tree to generate a final path.

[0037] In addition, the present application also provides a computer device comprising a processor and a memory, the memory is used for storing a computer executable program, the processor reads the computer executable program from the memory and executes, and the processor can realize the automatic driving path planning method in an unstructured environment based on a control line when executing the computer executable program.

[0038] Meanwhile, a computer readable storage medium can be provided, and the computer readable storage medium stores a computer program, and the computer program can realize the automatic driving path planning method in an unstructured environment based on a control line when being executed by a processor.

[0039] Compared with the prior art, the present application has at least the following beneficial effects: according to the distribution of obstacles in the environment, the search process ends after the newly expanded control line intersects with the target control line, ensuring the accuracy of the search path endpoint; in a large-scale scene, the search expansion can be carried out in large steps, ensuring the efficiency of the search; the path turning vector based on the method is less, ensuring the comfort of the path; and the experimental results in different complex environments show that the robustness is good. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 A method flowchart that can be implemented is provided.

[0041] Figure 2 A control line generation schematic diagram is provided.

[0042] Figure 3 A schematic diagram of different turning motion primitives is provided.

[0043] Figure 4 A CC-Turn generation schematic diagram is provided.

[0044] Figure 5 A schematic diagram of an intersection control line smooth turning path is provided.

[0045] Figure 6 A reachable intersection schematic diagram is provided.

[0046] Figure 7 A child control line generation schematic diagram is provided.

[0047] Figure 8 This is a diagram illustrating the path selection process.

[0048] Figure 9 The figures show comparative experiments of different methods in different scenarios. Detailed Implementation

[0049] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments: The autonomous driving path planning method based on control lines provided in this application has the following overall process: Figure 1 As shown, C-Line*Planner employs a search-based "OPEN+CLOSED" framework. OPEN and CLOSED are two linked lists, where the CLOSED list stores expanded control lines, and the OPEN list stores unexpanded control lines. However, unlike traditional algorithms that generate a state tree step-by-step through bottom-up exploration, C-Line*Planner constructs a driving strategy using a top-down approach. First, it generates a starting control line and a target control line. After obtaining the current control line, it continuously expands to generate new sub-control lines. Once a newly generated sub-control line can reach the target control line, the search process ends. By backtracking from the target control line, a skeleton tree with multiple interconnected control lines can be generated. Through permutation and combination, a path set can be obtained from the skeleton tree. C-Line*Planner aims to generate "natural" or "human-like" paths, therefore, it can select the most comfortable path from the path set as the final path.

[0050] An autonomous driving path planning method based on control lines defines the following basic concepts:

[0051] Control lines: Control lines It is a set of states with a fixed orientation that do not collide with obstacles, defined as:

[0052]

[0053] in Indicates passing through reference point P ref A directed line (x, y) with direction θ. This represents the area occupied by obstacles. When n = 1, 2, 3, the generated control lines are as follows: Figure 2 As shown.

[0054] Steering motion elements: such as Figure 3 As shown, the vehicle travels from point P0(x0,y0,θ0,κ0,υ0) on the control line to point P. f At this point, the curvature reaches its maximum value κ. max If a circular arc is used as the basic element for vehicle steering motion, then at the point P0 where the arc and the straight line are tangent, the rate of change of curvature c = ∞. To satisfy curvature continuity, as follows...Figure 3 As shown, this invention uses a spiral (with curvature κ proportional to arc length l) as another basic element for steering motion. Starting from the vehicle's current position P0, integration yields x, y, and θ.

[0055]

[0056] Where (x0,y0,θ0,κ0)=(0,0,0,0), c is the rate of change of curvature, l∈[0,l] max ], l max =k max / c.

[0057] CC-Turn: such as Figure 4 As shown, for path planning methods employing a "line-to-line" strategy, the vehicle travels from the current control line to the next control line, with both the starting and ending curvatures being zero. Based on changes in the vehicle's steering angle and whether gear shifting is required during travel, three motion primitives (control line, circular arc, and spiral) can be combined to construct different CC-Turns (Curvature Continue Turns). If gear shifting is not required during travel, a path is formed as shown below. Figure 4 (a) shows CC-In-Turn. If a gear shift is required during driving, it will form as shown in Figure (a). Figure 4 (b) shows the tangent point P between CC-Out-Turn, CC-Turn, and the control line. t It is determined by the angle Δθ between the two control lines.

[0058] Smooth turning path: For two intersecting control lines, such as Figure 5 As shown, the vehicle is moving from the current control line. Proceed to the next control line The vehicle's local working space can be divided into four distinct zones: I, II, III, and IV. Each zone corresponds to one CC-Turn and one C-Turn. The four different CC-Turns and C-Turns are linked to the current control line. and the next control line 16 tangent points are generated. Given a vehicle on the current control line... Position P in s The vehicle can first follow the current control line. Proceed to one of the tangent points P t1 Then use CC-Turn or C-Turn to turn to the next control line. The point of tangency P on t2 Finally, follow the next control line. Reach the target location P g .

[0059] Reachability and distance: As shown in Figure 6 When there are obstacles, it is difficult for two control lines to meet strict intersection, and the intersection point does not necessarily exist. As long as the reachable intersection is met, there is a CC-Turn or C-Turn without collision from one control line to another control line. The reachability R between two adjacent control lines is measured by the number of collision-free CC-Turn and C-Turn. The distance between two adjacent control lines is not unique, and the average curve distance CurveDis is used to measure,

[0060]

[0061] Where len i is the length of the i-th CC-Turn or C-Turn.

[0062] The application provides a control line-based automatic driving path planning method, which specifically comprises the following steps:

[0063] Step 1, initialization: the control line is a set of collision-free state sets with fixed orientation, and the starting control line init (x,y,θ) and the target control line x goal (x,y,θ) can be generated according to the starting node x and the target node x

[0064] Step 2, obtain the current control line: all sub-control lines are stored in the OPEN list, and the control line with the minimum evaluation function f value is selected as the current control line

[0065]

[0066] Step 3, sub-control line expansion: the current control line is taken as the parent control line, and the sub-control line can be generated through the defined action u succ , as shown in Figure 7 . The evaluation function f value is assigned to the sub-control line detected by the reachability, and the definition of f value is:

[0067]

[0068] Where g is the cost function of the distance from the starting control line, and h is the heuristic function of the distance from the target control line. The calculation of the cost function g is the cost of the current parent control line plus the cost between the current control line and the sub-control line:

[0069]

[0070] Since C-Line*Planner adopts line-to-line strategy, there are many CC-Turns or C-Turns between two adjacent reachable intersection of current control line and sub control line. The cost between current control line and sub control line is represented by average curve distance:

[0071]

[0072] where R denotes the reachability between two adjacent control lines, len i is the length of the ith CC-Turn or C-Turn.

[0073] The heuristic function h is defined as:

[0074]

[0075] where denotes the shortest straight line distance between current control line and target control line.

[0076] Finally, the newly generated sub control line is embedded into the OPEN list, and the parent control line which has finished expanding is embedded into the CLOSED list.

[0077] Step four, target control line backtracking: for traditional state tree based algorithm, when the target condition is satisfied, and backtracking from target node x goal to the start node x init , the final path can be obtained. However, for C-Line*Planner, when the newly expanded sub control line meets the end point reachable intersection condition with the target control line, the search process can be terminated, and from the target control line, the parent control lines are added into the final skeleton tree one by one.

[0078] Step five, final path selection: the generated skeleton tree is shown in Figure 8 , at the intersection of two control lines, there are still different collision free CC-Turns or C-Turns which can be selected to generate a large set of paths. Therefore, it is necessary to make further selection in the generated skeleton tree .

[0079] Suppose the number of reachable intersection in the skeleton tree is n, and R i is the reachability of the ith intersection. After permutation and combination, the path set

[0080]

[0081]

[0082] The number N of paths in the path set is:

[0083]

[0084] The path set can be rewritten as:

[0085]

[0086] From the path set Select a combination of the minimum number of shift points from the path set

[0087] Based on the inventive concept of the method, the application also provides a control line-based automatic driving unstructured environment path planning system for implementing the method, comprising an initialization module, a current control line acquisition module, a sub-control line extension module, a target control line backtracking module and a final path selection module; the initialization module is used for initializing a starting control line and a target control line;

[0088] The current control line acquisition module selects a control line with the minimum evaluation function f value from the sub-control lines as the current control line;

[0089] The sub-control line extension module takes the current control line as the parent control line, generates a sub-control line through the defined action u succ

[0090] The target control line backtracking module is used for backtracking the target control line to generate a final skeleton tree;

[0091] The final path selection module further selects a final path in the generated skeleton tree.

[0092] In addition, the application can also provide a computer device comprising a processor and a memory, the memory being used for storing a computer executable program, the processor reading part or all of the computer executable program from the memory and executing, and the processor being capable of realizing the control line-based automatic driving unstructured environment path planning method of the application when executing part or all of the computer executable program.

[0093] In another aspect, the application provides a computer readable storage medium, the computer readable storage medium storing a computer program, the computer program being capable of realizing the control line-based automatic driving unstructured environment path planning method of the application when being executed by a processor.

[0094] The computer device can be a notebook computer, a desktop computer, a vehicle-mounted computer or a workstation.

[0095] ​​​The processor can be a central processing unit (CPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), or a field-programmable gate array (FPGA).

[0096] For the memory described in the present application, it can be an internal storage unit of a notebook computer, a desktop computer, a vehicle-mounted computer or a workstation, such as a memory, a hard disk; or an external storage unit, such as a mobile hard disk, a flash card.

[0097] The computer-readable storage medium can include computer storage medium and communication medium. The computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information such as computer readable instructions, data structures, program modules or other data. The computer-readable storage medium can include read-only memory (ROM), random access memory (RAM), solid state disk (SSD) or optical disk, etc. Among them, the random access memory can include resistance random access memory (ReRAM) and dynamic random access memory (DRAM).

[0098] In order to prove the performance of the C-Line*Planner proposed in the present application, the C-Line*Planner is evaluated by batch simulation comparison test, and is compared with Hybrid A*(HA*) and CC-HA*. The position error of the termination condition of Hybrid A* and CC-HA* is set to 0.3m, the angle error is set to 5 degrees, and the maximum search time is set to 300ms. The C-Line*Planner uses C-Turn and CC-Turn respectively for ablation comparison experiment. In view of the fairness of the comparative experiment, Hybrid A* and CC-HA* do not use RS curve for analysis expansion. The length of the vehicle is 4.698m, the width is 1.942m, the wheelbase is 2.8m, and the maximum curvature is 0.3327. All simulation experiments in the present application are run on a computer equipped with Intel Core i7-10700 CPU and 16GB RAM.

[0099] Three scenarios are used for testing: (1) Parallel U-turn scenario without obstacles. The free space is set as 25m long and 8m wide. The start and goal poses of the vehicle are randomly generated, which are parallel to each other but in opposite directions. (2) Standard perpendicular parking scenario. The parking space is set as 6m deep and 2.5m wide. The initial drivable area is set as 6m wide and 20m long. The goal pose of parking is fixed, and the start pose is randomly generated in the initial drivable area of parking. (3) Random cluttered obstacle scenario, in which there are some narrow corridors. The start and goal poses of the vehicle are randomly generated, which are on the left and right sides of the narrow corridors, respectively. In each scenario, 100 sets of valid start and goal poses are sampled for path planning comparison test, and part of the test results are shown in Figure 9

[0100] To verify the performance of C-Line*Planner and the quality of the path, the evaluation indicators include success rate, search nodes / control lines number, time overhead, steering angle, number of gear shifting points and path length.

[0101] The search nodes / control lines number is the number of nodes (Hybrid A* and CC-HA*) or control lines (C-Line*) expanded in the search process, which is positively correlated with the time overhead. Fewer nodes or control lines mean higher computational efficiency. The steering angle (rad) is the absolute integral of the path curvature, which is a measure of the number of turns in the path. A path with fewer sharp turns or more straight sections is more comfortable than a path that requires a lot of turns along the way. The number of gear shifting points is the number of times the vehicle needs to shift gears during driving. A good path contains as few gear shifting points as possible.

[0102] Table 1 Summary of algorithm performance and path quality performance of 100 random tests.

[0103]

[0104]

[0105] The results of 100 random tests in three representative scenarios are shown in Table 1. The performance of Hybrid A*, CC-HA* and C-Line*Planner is compared and analyzed from two aspects of path planning algorithm performance and planned path quality.

[0106] ​The results of the algorithm performance involve three indicators, i.e. success rate, search nodes / control lines number and time cost. It can be seen that under the limit of maximum search time, the success rates of Hybrid A* in the three experimental scenarios are 69.5%, 100% and 73.5% respectively; the time cost of C-Line*+C-Turn in the three scenarios is much lower than 300s, so the success rates are 100%, 100% and 96.55% respectively. In the parallel turning scenario, CC-HA* and C-Line*+CC-Turn cannot find a collision-free path, and the success rates of other scenarios are lower than Hybrid A* and CC-Line*+CC-Turn, which shows that although the introduction of spiral turning motion primitives increases the smoothness of the path, it reduces the success rate of the algorithm.

[0107] It can be seen that the number of search nodes / control lines and the time cost are positively correlated. The number of control lines generated by C-Line* is much smaller than the number of nodes generated by Hybrid A* and CC-HA*, and the search time of C-Line* in the three scenarios is only 4.898%, 15.084% and 0.928% of that of Hybrid A* respectively, which means that the C-line* planner using the line-to-line expansion strategy has a longer vision and higher computational efficiency.

[0108] The results of path quality involve three other indicators, i.e. turning amount, path length and shift point number. The average path length of C-Line* in the three scenarios is 309.65%, 98.083% and 85.228% of that of Hybrid A* respectively, and the average shift point number is 99.8%, 127.6% and 100% of that of Hybrid A* respectively. The average turning amount is 88.661%, 59.6% and 22.84% of that of Hybrid A* respectively, which means that C-Line* planner produces the most comfortable and most "human-like" path in terms of turning amount, and the path length and shift point number are basically the same as Hybrid A* except in the parallel turning scenario. From Figure 9 In the cluttered obstacle scenario in Fig. 6, it can be seen that the path planned by C-Line* is very "straight", while the paths planned by Hybrid A* and CC-HA* are very "winding".

[0109] More importantly, accuracy is measured by the position error and direction error of the path endpoint. Since the C-Line* algorithm is always reachable with the target control line, there is no error in the path endpoint planned by the C-Line* algorithm. Due to the existence of search termination conditions, Hybrid A* and CC-HA* have unavoidable systematic errors at the path endpoint.

[0110] In summary, considering the algorithm performance, path quality and target accuracy, C-Line*Planner greatly improves the scene robustness, computational efficiency and path quality compared with Hybrid A*and CC-HA*. The paths generated by C-Line*+CC-Turn and CC-HA*are smooth enough for motion controller to track. The curvature discontinuity of Hybrid A*and C-Line*+C-Turn can be solved by adding an optimized post-processing end.

[0111] In summary, the present application provides a new control line based autonomous driving path planning algorithm, called C-Line*Planner(Control-Line Guided Path Planner). Control line(C-Line) is a set of states with fixed direction, under the "control" of two adjacent control lines, different motion primitives are spliced to generate a smooth turning path, which makes the two control lines reachable for autonomous vehicles. In the first stage, the starting control line and the target control line are generated first, and then new reachable sub-control lines are constantly expanded. Once the current control line can reach the target control line, the search process ends immediately, so that a skeleton tree with multiple connected control lines can be constructed. In the second stage, through permutation and combination, a path set can be obtained from the skeleton tree. The goal of C-Line*planner is to generate a "natural" or "human-like" path, so a more comfortable path can be selected from the path set as the final path.

Claims

1. A path planning method for unstructured environments in autonomous driving based on control lines, characterized in that, Specifically, the steps include the following: S1, Initialization, initialize the start control line and target control line; S2, Obtain the current control line, and select the evaluation function from all sub-control lines. The control line with the smallest value is taken as the current control line; the evaluation function is... The value is assigned to the sub-control line that passes the reachability test, and the evaluation function is evaluated. The value is defined as: in It is the cost function of the distance sub-control line. It is the heuristic function for the distance to the target control line. The cost function g is calculated as the cost of the current parent control line plus the cost between the current control line and the child control line. Between two adjacent, reachable, intersecting current control lines and sub-control lines, there are many CC-Turns or C-Turns. The average curve distance is used to represent the cost between the current control line and the sub-control line. Where R represents the reachability between two adjacent control lines. It is the first The length of a CC-Turn or C-Turn; Heuristic function for: in This represents the shortest straight-line distance between the current control line and the target control line; S3, Sub-control line extension, uses the current control line as the parent control line, through defined actions. Generate sub-control lines; S4, target control line backtracking, generating the final skeleton tree; S5, Final Path Selection: Further selection is performed in the generated skeleton tree to generate the final path.

2. The method for unstructured environment path planning for autonomous driving based on control lines according to claim 1, characterized in that, In S1, during initialization, the control lines are a set of states with fixed orientations that do not collide with obstacles, based on the vehicle's starting node. and target node Generate start control line and target control line .

3. The method for unstructured environment path planning for autonomous driving based on control lines according to claim 1, characterized in that, When a sub-control line is extended, the current control line is used. As the parent control line, the child control line By defined actions Generate sub-control lines through reachability detection and assign them to the evaluation function. Value, newly generated control line Embedded into OPEN In the linked list, the parent control line that has been expanded Embedded into CLOSED In a linked list.

4. The method for unstructured environment path planning for autonomous driving based on control lines according to claim 1, characterized in that, In S4, when the target control line is backtracked, the latest extended sub-control line... With the target control line Satisfying the reachability and intersection conditions, reachability The search process then ends, starting with the target control line, and parent control lines are added sequentially to the final skeleton tree. middle.

5. The method for unstructured environment path planning for autonomous driving based on control lines according to claim 1, characterized in that, In S5, during the final path selection, the path set is obtained by arranging and combining the path sets at the intersections of the two control lines in the generated skeleton tree. , From path set Choose the combination with the minimum number of shift points as the final path.

6. The method for unstructured environment path planning for autonomous driving based on control lines according to claim 1, characterized in that, Throughout the vehicle's steering motion, circular arcs and spirals are used as the basic elements of the steering motion.

7. A path planning system for unstructured environments for autonomous driving based on control lines, characterized in that, It includes an initialization module, a current control line acquisition module, a sub-control line expansion module, a target control line backtracking module, and a final path selection module; the initialization module is used to initialize the starting control line and the target control line; The current control line acquisition module selects the evaluation function from the sub-control lines. The control line with the smallest value is taken as the current control line; the evaluation function is... The value is assigned to the sub-control line that passes the reachability test, and the evaluation function is evaluated. The value is defined as: in It is the cost function of the distance sub-control line. It is the heuristic function for the distance to the target control line. The cost function g is calculated as the cost of the current parent control line plus the cost between the current control line and the child control line. Between two adjacent, reachable, intersecting current control lines and sub-control lines, there are many CC-Turns or C-Turns. The average curve distance is used to represent the cost between the current control line and the sub-control line. Where R represents the reachability between two adjacent control lines. It is the first The length of a CC-Turn or C-Turn; Heuristic function for: in This represents the shortest straight-line distance between the current control line and the target control line; The child control line extension module uses the current control line as the parent control line and defines actions. Generate sub-control lines; The target control line backtracking module is used to backtrack the target control line and generate the final skeleton tree; The final path selection module performs further selections in the generated skeleton tree to generate the final path.

8. A computer device, characterized in that, It includes a processor and a memory, the memory being used to store a computer-executable program, the processor reading the computer-executable program from the memory and executing it, and the processor executing the computer-executable program is capable of implementing the unstructured environment path planning method for autonomous driving based on control lines as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, A computer-readable storage medium stores a computer program that, when executed by a processor, enables the implementation of the control line-based unstructured environment path planning method for autonomous driving as described in any one of claims 1 to 6.

Images