A method and system for path planning in cartesian space with drivable corridor constraints

Abstract

The application relates to a kind of Cartesian space path planning methods and systems of travel corridor constraint, the method comprises: S1. Discrete sampling is carried out in Frenet space, and Frenet rough path is generated by dynamic programming search;S2. Based on Frenet rough path, obstacle information and road boundary, construct travel corridor;S3. The Frenet travel corridor is converted to the ordered convex polygon corridor under Cartesian space;S4. Under the constraint of the ordered convex polygon corridor sequence, the anchor point sequence is determined by tangent line cone angle interval search.The application not only effectively avoids the path swing and curvature mutation caused by reference line distortion, but also improves the path smoothness.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and in particular to a Cartesian space path planning method and system with drivable corridor constraints. Background Technology

[0002] Existing autonomous driving path planning methods mostly employ Frenet space based on reference lines. This method uses the arc length *s* of the reference line to represent the longitudinal position and the lateral offset *l* relative to the reference line to represent the lateral position, thus transforming road boundary and obstacle constraints into a relatively regular search problem, and is therefore widely used for online path planning. However, the planning results in Frenet space are highly dependent on the geometry of the reference line. When map noise, road stitching errors, abrupt curvature changes, or local obstacle avoidance cause reference line distortion, even if the lateral offset *l* remains continuous and smooth in Frenet space, the path may still exhibit unnecessary swaying, abrupt curvature changes, or detours after conversion to Cartesian space, affecting vehicle tracking stability and ride comfort. Path planning in Cartesian space directly generates paths based on the real planar positions of the vehicle, obstacles, and road boundaries, reducing the dependence on the geometry of the reference line.

[0003] However, when A*, hybrid A*, state grid, or sampling search are directly expanded across the entire two-dimensional Cartesian space, a large number of candidate nodes need to be maintained, and reachability judgments and collision detections need to be repeatedly performed. The search time increases significantly with the map range, resolution, and number of obstacles, making it difficult to guarantee the real-time performance required for high-frequency rolling planning. Chinese invention patent application publication number CN115730756A discloses a phased trajectory planning method for autonomous vehicles. In the search phase, sampling is performed in the Frenet coordinate system, and a reference trajectory is obtained through dynamic planning and expanded into a drivable path. In the optimization phase, the system is transformed to the Cartesian coordinate system, and the trajectory is output using the ILQR algorithm. However, the distortion of the reference line is propagated to the final path. Chinese invention patent CN112612267B discloses an autonomous driving path planning method. It obtains a heuristic function through deep reinforcement learning to guide the Hybrid A* search, reducing redundant exploration, and then generates a smooth trajectory through nonlinear optimization and nonparametric interpolation. However, model reasoning needs to be completed before the search, and candidate trees need to be maintained node by node, heuristic functions need to be calculated and collision detection needs to be repeated during the search. This makes it difficult to meet the requirements of high-frequency rolling planning for low latency and deterministic computation time, and the real-time performance is still insufficient.

[0004] In the prior art, Chinese patent (publication number: CN115730756A) discloses a phased trajectory planning method for autonomous vehicles. In the search phase, sampling is performed in the Frenet coordinate system, and a fifth-order polynomial is used to connect the sampling points. Dynamic programming is then used to search for a reference trajectory and construct a drivable path. In the optimization phase, a vehicle kinematic model is established, a cost function and boundary constraints are constructed, and the ILQR algorithm is used for spatiotemporal optimization to output the optimal trajectory. However, in this scheme, the Frenet spatial planning method uses the reference line as the geometric benchmark of the final path, and local distortions of the reference line can be propagated to the planned path.

[0005] In the prior art, Chinese patent (publication number: CN112612267B) discloses a path planning method and apparatus for autonomous driving, including: acquiring a grid space image; obtaining the state value of the grid space image under the current environment through a deep reinforcement learning algorithm; obtaining the distance from the starting point to the ending point by applying an autonomous driving motion planning curve to the grid space image; obtaining the heuristic function value corresponding to each grid node based on the state value under the current environment and the distance from the starting point to the ending point; obtaining the driving motion path based on the heuristic function value corresponding to each grid node; and optimizing the driving motion path to obtain a smooth trajectory. However, the existing Cartesian space global search in this scheme requires expanding a large number of candidate nodes and frequently performing collision detection, or relies on a learning model to provide heuristic information for the search, resulting in a large online computational load and insufficient real-time performance. Summary of the Invention

[0006] In view of the shortcomings of the prior art, the present invention provides a Cartesian space path planning method and system with drivable corridor constraints, which not only effectively avoids path swaying and curvature abrupt changes caused by reference line distortion, but also improves path smoothness.

[0007] To achieve the above and other related objectives, the present invention provides the following technical solution: A Cartesian space path planning method with drivable corridor constraints, the method comprising: S1. Perform discrete sampling in the Frenet space and generate Frenet coarse paths through dynamic programming search; S2. Construct a drivable corridor based on Frenet coarse path, obstacle information, and road boundaries; S3. Convert the Frenet drivable corridor into an ordered convex polygon corridor in Cartesian space; S4. Under the constraint of the ordered convex polygon corridor sequence, the anchor point sequence is determined by searching the tangent cone angle interval; S5. Perform interpolation on the generated path and output a coarse Cartesian path from the start point to the end point.

[0008] Further, in step S1, the dynamic programming search to generate the Frenet coarse path includes: S11. Let the road reference line where the autonomous vehicle is located have an arc length s∈[0, s] in the Frenet coordinate system. max The vertical position is represented by ], and the horizontal offset l represents the horizontal position relative to the reference line. The reference line is discretized along the longitudinal direction at a fixed sampling interval Δs to obtain a set of longitudinal sampling points. , Where K is the total number of vertical sampling points; S12. At each longitudinal sampling point s i At this location, sampling is performed along the lateral direction at a fixed sampling interval Δl to obtain a set of lateral offset sampled values. , Where M i Let be the total number of lateral sampling points at the i-th layer. Each sampling point must satisfy the road boundary constraints, i.e. , l min (s i ) and l max (s i ) represent the road boundaries at s i Minimum and maximum permissible lateral offsets of the location; S13. Starting from the point The root node is the endpoint. For the target node, a dynamic programming algorithm is used to search for the minimum cost path in the graph structure formed by the sampled points. , in, For a candidate path from the origin to the destination, c(l) i ,l i+1 ) represents the path segment cost function between sampling points in adjacent layers. The path obtained from the search is the Frenet coarse path, represented as a sequence. .

[0009] Further, in step S2, constructing the drivable corridor includes: S21. Obtain the projected position information of the obstacle in the Frenet coordinate system. Let the vertical interval occupied by the q-th obstacle be... The horizontally occupied area is Define the road boundary constraint function as follows: , in, Let be the lateral coordinate of the left boundary of the road at the longitudinal position 's'. Let be the lateral coordinate of the right boundary of the road at the longitudinal position s; S22. At each longitudinal position s i At this point, the drivable lateral section is defined as the largest continuous section that satisfies both obstacle avoidance constraints and road boundary constraints. , in, The left boundary of the drivable road. This is the right boundary of the drivable road. S23. Using Frenet coarse path To select a reference point, at each longitudinal position s i Select the location containing the coarse path point. The feasible intervals are used as the corresponding drivable lateral intervals for that layer. By combining the drivable lateral intervals at each longitudinal position, a complete Frenet drivable corridor is obtained. .

[0010] Furthermore, the left boundary of the drivable road for, , Where q represents the vertical position s i There are all obstacles occupying the horizontal position at point d. safe For safe obstacle avoidance distance.

[0011] Furthermore, the right boundary of the drivable road for, , Where q represents the vertical position s i There are all obstacles occupying the horizontal position at point d. safe For safe obstacle avoidance distance.

[0012] Further, in step S3, converting the Frenet drivable corridor into an ordered convex polygon corridor in Cartesian space includes: S31. Suppose that the Frenet drivable corridor consists of several consecutive drivable sections, and each drivable section is represented in the Frenet coordinate system as follows: , Where s is the longitudinal arc length along the reference line, and l is the lateral offset relative to the reference line. j min and l j max These are the left and right boundaries of the j-th drivable section, respectively. j min and s jmax These are the rear and front boundaries of the j-th drivable section, respectively; S32. Obtain the position of the reference point at the arc length s of the reference line. and heading angle Construct the unit normal vector And convert Frenet space points to Cartesian space; S33. For each R Freret After transformation to Cartesian space using T(s,l), an ordered sequence of convex polygon corridors is generated. .

[0013] Furthermore, the conversion of Frenet space points to Cartesian space is... , Where s is the longitudinal arc length along the reference line, l is the lateral offset relative to the reference line, r(s) is the position of the reference point, and q r (s) is the heading angle.

[0014] Further, in step S4, determining the anchor point sequence through tangent cone angle interval search includes: S41. Define the current anchor point set A = {A0, A1, ..., A...} m}, the initial anchor point A0 is the path starting point p start For any anchor point A k And for the i-th polygon Pi, define the visible tangent cone C(A) from to . k ,P i )for, , The left boundary q of the tangent cone i L and right boundary q i R From A respectively k To P i The direction angles of the two tangents drawn from the nearest or farthest edge are determined, and the corresponding tangent points are recorded as Q. i =(Q i L Q i R ); S42. Define the current tangent cone as the visible angle interval. The initial state is , indicating full-width visibility, for the current anchor point A k Starting from the polygon at layer i=k+1, subsequent polygons are sequentially included in the visibility scan. For each polygon P... i Calculate the tangent cone C(A) k,P i And take the intersection with the current tangent cone. , If satisfied, , Where ε is the preset cross tolerance, it is determined that the tangent cone has not collapsed, and the current tangent cone is updated. , Continue scanning the next polygon; if the conditions are met, , The tangent cone is then determined to have collapsed, indicating that the current anchor point A... k Since the i-th layer polygon cannot be directly observed, a new anchor point A is selected based on the boundary type that caused the collapse. k+1 If the collapse is triggered by the right boundary update, i.e., the candidate right boundary angle q i R Less than the current left boundary angle q new L Then set the current left boundary tangent point as the new anchor point. A k+1 =Q j L , Where j is the index of the polygon containing the current left boundary; S43. If the collapse is triggered by the left boundary update, i.e., the candidate left boundary angle q i L Greater than the current right boundary angle q new L Then set the tangent point corresponding to the current right boundary as the new anchor point. A k+1 =Q j R Then, it was assumed that the new current anchor point A k+1 Update the scan start position to i, that is, start the scan again from the current layer where the collapse occurred, and reset the tangent cone to a full-angle domain. , Continue performing the above scanning process, repeating the process until the last layer of polygon P is scanned. N-1 If the endpoint p goal Located at the final tangent cone F m Inside, , Then p goal As the final anchor point, it is added to the anchor point set to obtain the complete anchor point sequence. .

[0015] Further, in step S5, the interpolation performed on the generated path to output a coarse Cartesian path from the starting point to the ending point includes: S51. For each pair of adjacent anchor points in the anchor point sequence A path segment connecting two points is generated using linear interpolation. Let the path segment be parameterized as follows: Its expression is, , t∈[0,1], Where, parameter t is the normalized interpolation coefficient, and when t=0, g j (0)=A j The path segment begins at anchor point A. j When t=1, the path segment terminates at anchor point A. j+1 When t∈(0,1), g j (t) indicates that it is located at anchor point A. j With A j+1 A series of intermediate interpolation points on the line; S52. Perform the linear interpolation on each adjacent anchor point pair in sequence, and merge all interpolated path segments with the original anchor point sequence to obtain a complete set of coarse path points in Cartesian space. , That is, the coarse path is formed by all anchor points and the straight-line interpolation points between adjacent anchor points, forming a path from the starting point A0 to the ending point A0. m Complete piecewise linear continuous path , Interpolation points can be selected on the path segment using equal spacing or equal arc length.

[0016] To achieve the above and other related objectives, the present invention also provides a drivable corridor-constrained Cartesian space path planning system, including a computer device programmed or configured to perform the steps of the drivable corridor-constrained Cartesian space path planning method.

[0017] The present invention has the following positive effects: In the path generation stage, this invention directly determines the path geometry based on the start point, end point, and corridor boundaries in XY space, eliminating the need for reference lines as shape benchmarks. This effectively avoids path swaying and curvature abrupt changes caused by reference line distortion, thus improving path smoothness. Simultaneously, this invention transforms large-scale two-dimensional path search into a sequential scanning process of corridor sequences, eliminating the need to maintain a large-scale search tree or expand numerous nodes, significantly reducing computational overhead and meeting the real-time requirements of autonomous driving path planning. Attached Figure Description

[0018] Figure 1 This is a flowchart of the overall path planning in Cartesian space based on drivable corridor constraints according to the present invention. Figure 2 This is a schematic diagram (a) of the point selection and path generation process based on the tangent cone angle interval of the present invention; Figure 3 This is a schematic diagram (b) of the point selection and path generation process based on the tangent cone angle interval of the present invention; Figure 4 This is a schematic diagram (c) of the point selection and path generation process based on the tangent cone angle interval of the present invention; Figure 5 This is a schematic diagram (d) of the point selection and path generation process based on the tangent cone angle interval of the present invention. Detailed Implementation

[0019] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0020] Example: Figure 1 As shown, a Cartesian space path planning method with drivable corridor constraints is described, the method comprising: S1. Perform discrete sampling in the Frenet space and generate Frenet coarse paths through dynamic programming search; S2. Construct a drivable corridor based on Frenet coarse path, obstacle information, and road boundaries; S3. Convert the Frenet drivable corridor into an ordered convex polygon corridor in Cartesian space; S4. Under the constraint of the ordered convex polygon corridor sequence, the anchor point sequence is determined by searching the tangent cone angle interval; S5. Perform interpolation on the generated path and output a coarse Cartesian path from the start point to the end point.

[0021] In this embodiment, in step S1, the dynamic programming search to generate the Frenet coarse path includes: S11. Let the road reference line where the autonomous vehicle is located have an arc length s∈[0, s] in the Frenet coordinate system. max The vertical position is represented by ], and the horizontal offset l represents the horizontal position relative to the reference line. The reference line is discretized along the longitudinal direction at a fixed sampling interval Δs to obtain a set of longitudinal sampling points. , Where K is the total number of vertical sampling points; S12. At each longitudinal sampling point s iAt this location, sampling is performed along the lateral direction at a fixed sampling interval Δl to obtain a set of lateral offset sampled values. , Where M i Let be the total number of lateral sampling points at the i-th layer. Each sampling point must satisfy the road boundary constraints, i.e. , l min (s i ) and l max (s i ) represent the road boundaries at s i Minimum and maximum permissible lateral offsets of the location; S13. Starting from the point The root node is the endpoint. For the target node, a dynamic programming algorithm is used to search for the minimum cost path in the graph structure formed by the sampled points. , in, For a candidate path from the origin to the destination, c(l) i ,l i+1 ) represents the path segment cost function between sampling points in adjacent layers. The path obtained from the search is the Frenet coarse path, represented as a sequence. .

[0022] In this embodiment, in step S2, constructing the drivable corridor includes: S21. Obtain the projected position information of the obstacle in the Frenet coordinate system. Let the vertical interval occupied by the q-th obstacle be... The horizontally occupied area is Define the road boundary constraint function as follows: , in, Let be the lateral coordinate of the left boundary of the road at the longitudinal position 's'. Let be the lateral coordinate of the right boundary of the road at the longitudinal position s; S22. At each longitudinal position s i At this point, the drivable lateral section is defined as the largest continuous section that satisfies both obstacle avoidance constraints and road boundary constraints. , in, The left boundary of the drivable road. This is the right boundary of the drivable road. S23. Using Frenet coarse path To select a reference point, at each longitudinal position s iSelect the location containing the coarse path point. The feasible intervals are used as the corresponding drivable lateral intervals for that layer. By combining the drivable lateral intervals at each longitudinal position, a complete Frenet drivable corridor is obtained. .

[0023] In this embodiment, the left boundary of the drivable road for, , Where q represents the vertical position s i There are all obstacles occupying the horizontal position at point d. safe For safe obstacle avoidance distance.

[0024] In this embodiment, the right boundary of the drivable road for, , Where q represents the vertical position s i There are all obstacles occupying the horizontal position at point d. safe For safe obstacle avoidance distance.

[0025] In this embodiment, step S3, converting the Frenet drivable corridor to an ordered convex polygon corridor in Cartesian space, includes: S31. Suppose that the Frenet drivable corridor consists of several consecutive drivable sections, and each drivable section is represented in the Frenet coordinate system as follows: , Where s is the longitudinal arc length along the reference line, and l is the lateral offset relative to the reference line. j min and l j max These are the left and right boundaries of the j-th drivable section, respectively. j min and s j max These are the rear and front boundaries of the j-th drivable section, respectively; S32. Obtain the position of the reference point at the arc length s of the reference line. and heading angle Construct the unit normal vector And convert Frenet space points to Cartesian space; S33. For each R Freret After transformation to Cartesian space using T(s,l), an ordered sequence of convex polygon corridors is generated. .

[0026] In this embodiment, the conversion of Frenet space points to Cartesian space is... , Where s is the longitudinal arc length along the reference line, l is the lateral offset relative to the reference line, r(s) is the position of the reference point, and q r (s) is the heading angle.

[0027] In this embodiment, in step S4, determining the anchor point sequence through tangent cone angle interval search includes: S41. Define the current anchor point set A = {A0, A1, ..., A...} m}, the initial anchor point A0 is the path starting point p start For any anchor point A k And for the i-th polygon Pi, define the visible tangent cone C(A) from to . k ,P i )for, , The left boundary q of the tangent cone i L and right boundary q i R From A respectively k To P i The direction angles of the two tangents drawn from the nearest or farthest edge are determined, and the corresponding tangent points are recorded as Q. i =(Q i L Q i R ); S42. Define the current tangent cone as the visible angle interval. The initial state is , indicating full-width visibility, for the current anchor point A k Starting from the polygon at layer i=k+1, subsequent polygons are sequentially included in the visibility scan. For each polygon P... i Calculate the tangent cone C(A) k ,P i And take the intersection with the current tangent cone. , If satisfied, , Where ε is the preset cross tolerance, it is determined that the tangent cone has not collapsed, and the current tangent cone is updated. , Continue scanning the next polygon; if the conditions are met, , The tangent cone is then determined to have collapsed, indicating that the current anchor point A... k Since the i-th layer polygon cannot be directly observed, a new anchor point A is selected based on the boundary type that caused the collapse. k+1 If the collapse is triggered by the right boundary update, i.e., the candidate right boundary angle q i R Less than the current left boundary angle q new L Then set the current left boundary tangent point as the new anchor point. A k+1 =Q j L , Where j is the index of the polygon containing the current left boundary; S43. If the collapse is triggered by the left boundary update, i.e., the candidate left boundary angle q i L Greater than the current right boundary angle q new L Then set the tangent point corresponding to the current right boundary as the new anchor point. A k+1 =Q j R Then, it was assumed that the new current anchor point A k+1 Update the scan start position to i, that is, start the scan again from the current layer where the collapse occurred, and reset the tangent cone to a full-angle domain. , Continue performing the above scanning process, repeating the process until the last layer of polygon P is scanned. N-1 If the endpoint p goal Located at the final tangent cone F m Inside, , Then p goal As the final anchor point, it is added to the anchor point set to obtain the complete anchor point sequence. .

[0028] In this embodiment, as Figure 2 or Figure 3 or Figure 4 or Figure 5 In step S5, the process of interpolating the generated path and outputting a coarse Cartesian path from the starting point to the ending point includes: S51. For each pair of adjacent anchor points in the anchor point sequence A path segment connecting two points is generated using linear interpolation. Let the path segment be parameterized as follows: Its expression is, , t∈[0,1], Where, parameter t is the normalized interpolation coefficient, and when t=0, gj (0)=A j The path segment begins at anchor point A. j When t=1, the path segment terminates at anchor point A. j+1 When t∈(0,1), g j (t) indicates that it is located at anchor point A. j With A j+1 A series of intermediate interpolation points on the line; S52. Perform the linear interpolation on each adjacent anchor point pair in sequence, and merge all interpolated path segments with the original anchor point sequence to obtain a complete set of coarse path points in Cartesian space. , That is, the coarse path is formed by all anchor points and the straight-line interpolation points between adjacent anchor points, forming a path from the starting point A0 to the ending point A0. m Complete piecewise linear continuous path , Interpolation points can be selected on the path segment using equal spacing or equal arc length.

[0029] In this embodiment, the present invention also provides a drivable corridor-constrained Cartesian space path planning system, including a computer device programmed or configured to perform the steps of the drivable corridor-constrained Cartesian space path planning method.

[0030] In summary, this invention not only effectively avoids path swaying and curvature abrupt changes caused by reference line distortion, but also improves path smoothness.

[0031] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A Cartesian space path planning method with drivable corridor constraints, characterized in that, The method includes: S1. Perform discrete sampling in the Frenet space and generate Frenet coarse paths through dynamic programming search; S2. Construct a drivable corridor based on Frenet coarse path, obstacle information, and road boundaries; S3. Convert the Frenet drivable corridor into an ordered convex polygon corridor in Cartesian space; S4. Under the constraint of the ordered convex polygon corridor sequence, the anchor point sequence is determined by searching the tangent cone angle interval; S5. Perform interpolation on the generated path and output a coarse Cartesian path from the start point to the end point.

2. The Cartesian space path planning method with drivable corridor constraints according to claim 1, characterized in that, In step S1, the dynamic programming search to generate the Frenet coarse path includes: S11. Let the road reference line where the autonomous vehicle is located have an arc length s∈[0, s] in the Frenet coordinate system. max The vertical position is represented by ], and the horizontal offset l represents the horizontal position relative to the reference line. The reference line is discretized along the longitudinal direction at a fixed sampling interval Δs to obtain a set of longitudinal sampling points. , Where K is the total number of longitudinal sampling points; S12. At each longitudinal sampling point s i At this location, sampling is performed along the lateral direction at a fixed sampling interval Δl to obtain a set of lateral offset sampled values. ， Where M i Let be the total number of lateral sampling points at the i-th layer. Each sampling point must satisfy the road boundary constraints, i.e. , l min (s i ) and l max (s i ) represent the road boundaries at s i Minimum and maximum permissible lateral offsets of the location; S13. Starting from the point The root node is the endpoint. For the target node, a dynamic programming algorithm is used to search for the minimum cost path in the graph structure formed by the sampled points. ， in, For a candidate path from the starting point to the ending point, c(l) i ,l i+1 ) represents the path segment cost function between sampling points in adjacent layers. The path obtained from the search is the Frenet coarse path, represented as a sequence. 。 3. The Cartesian space path planning method with drivable corridor constraints according to claim 1, characterized in that, In step S2, constructing the drivable corridor includes: S21. Obtain the projected position information of the obstacle in the Frenet coordinate system. Let the vertical interval occupied by the q-th obstacle be... The horizontally occupied area is Define the road boundary constraint function as follows: ， in, Let be the lateral coordinate of the left boundary of the road at the longitudinal position 's'. Let be the lateral coordinate of the right boundary of the road at the longitudinal position s; S22. At each longitudinal position s i At this point, the drivable lateral section is defined as the largest continuous section that satisfies both obstacle avoidance constraints and road boundary constraints. ， in, The left boundary of the drivable road. This is the right boundary of the drivable road. S23. Using Frenet coarse path To select a reference point, at each longitudinal position s i Select the location containing the coarse path point. The feasible intervals are used as the corresponding drivable lateral intervals for that layer. By combining the drivable lateral intervals at each longitudinal position, a complete Frenet drivable corridor is obtained. 。 4. The Cartesian space path planning method with drivable corridor constraints according to claim 3, characterized in that: The left boundary of the drivable road for, ， Where q represents the vertical position s i There are all obstacles occupying the horizontal position at point d. safe For safe obstacle avoidance distance.

5. The Cartesian space path planning method with drivable corridor constraints according to claim 3, characterized in that: The right boundary of the drivable road for, , Where q represents the vertical position s i There are all obstacles occupying the horizontal position at point d. safe For safe obstacle avoidance distance.

6. The Cartesian space path planning method with drivable corridor constraints according to claim 1, characterized in that, In step S3, converting the Frenet drivable corridor into an ordered convex polygon corridor in Cartesian space includes: S31. Suppose that the Frenet drivable corridor consists of several consecutive drivable sections, and each drivable section is represented in the Frenet coordinate system as follows: ， Where s is the longitudinal arc length along the reference line, and l is the lateral offset relative to the reference line. j min and l j max These are the left and right boundaries of the j-th drivable section, respectively. j min and s j max These are the rear and front boundaries of the j-th drivable section, respectively; S32. Obtain the position of the reference point at the arc length s of the reference line. and heading angle Construct the unit normal vector And convert Frenet space points to Cartesian space; S33. For each R Freret After transformation to Cartesian space using T(s,l), an ordered sequence of convex polygon corridors is generated. 。 7. The Cartesian space path planning method with drivable corridor constraints according to claim 6, characterized in that: The process of converting Frenet space points to Cartesian space is as follows: ， Where s is the longitudinal arc length along the reference line, l is the lateral offset relative to the reference line, r(s) is the position of the reference point, and q r (s) is the heading angle.

8. The Cartesian space path planning method with drivable corridor constraints according to claim 1, characterized in that, In step S4, determining the anchor point sequence through tangent cone angle interval search includes: S41. Define the current anchor point set A = {A0, A1, ..., A...} m }, the initial anchor point A0 is the path starting point p start For any anchor point A k And for the i-th polygon Pi, define the visible tangent cone C(A) from to . k ,P i )for, , The left boundary q of the tangent cone i L and right boundary q i R From A respectively k To P i The direction angles of the two tangents drawn from the nearest or farthest edge are determined, and the corresponding tangent points are recorded as Q. i =(Q i L Q i R ); S42. Define the current tangent cone as the visible angle interval. The initial state is , indicating full-width visibility, for the current anchor point A k Starting from the polygon at layer i=k+1, subsequent polygons are sequentially included in the visibility scan. For each polygon P... i Calculate the tangent cone C(A) k ,P i And take the intersection with the current tangent cone. , If satisfied, , Where ε is the preset cross tolerance, it is determined that the tangent cone has not collapsed, and the current tangent cone is updated. ， Continue scanning the next polygon; if the conditions are met, ， The tangent cone is then determined to have collapsed, indicating that the current anchor point A... k Since the i-th layer polygon cannot be directly observed, a new anchor point A is selected based on the boundary type that caused the collapse. k+1 If the collapse is triggered by the right boundary update, i.e., the candidate right boundary angle q i R Less than the current left boundary angle q new L Then set the current left boundary tangent point as the new anchor point. A k+1 =Q j L , Where j is the index of the polygon containing the current left boundary; S43. If the collapse is triggered by the left boundary update, i.e., the candidate left boundary angle q i L Greater than the current right boundary angle q new L Then set the tangent point corresponding to the current right boundary as the new anchor point. A k+1 =Q j R Then, it was assumed that the new current anchor point A k+1 Update the scan start position to i, that is, start the scan again from the current layer where the collapse occurred, and reset the tangent cone to a full-angle domain. And continue performing the above scanning process, repeating the process until the last layer of polygon P is scanned. N-1 If the endpoint p goal Located at the final tangent cone F m Inside, , Then p goal As the final anchor point, it is added to the anchor point set to obtain the complete anchor point sequence. .

9. The Cartesian space path planning method with drivable corridor constraints according to claim 1, characterized in that, In step S5, interpolating the generated path and outputting a coarse Cartesian path from the starting point to the ending point includes: S51. For each pair of adjacent anchor points in the anchor point sequence A path segment connecting two points is generated using linear interpolation. Let the path segment be parameterized as follows: Its expression is, ，t∈[0,1]， Where, parameter t is the normalized interpolation coefficient, and when t=0, g j (0)=A j The path segment begins at anchor point A. j When t=1, the path segment terminates at anchor point A. j+1 When t∈(0,1), g j (t) indicates that it is located at anchor point A. j With A j+1 A series of intermediate interpolation points on the line; S52. Perform the linear interpolation on each adjacent anchor point pair in sequence, and merge all interpolated path segments with the original anchor point sequence to obtain a complete set of coarse path points in Cartesian space. , That is, the coarse path is formed by all anchor points and the straight-line interpolation points between adjacent anchor points, forming a path from the starting point A0 to the ending point A0. m Complete piecewise linear continuous path , Interpolation points can be selected on the path segment using equal spacing or equal arc length.

10. A Cartesian spatial path planning system with drivable corridor constraints, comprising a computer device, characterized in that, The computer device is programmed or configured to perform the steps of the Cartesian spatial path planning method with drivable corridor constraints as described in any one of claims 1 to 9.

Images