CN112706760A - Unmanned parking path planning method for special road scene - Google Patents

Abstract

The invention discloses a method for planning an unmanned parking path for a special road scene, which divides a road space from a vehicle starting point to a parking terminal point into a plurality of space pieces by taking a transit stop point as a demarcation point, and can improve the solving speed of an algorithm; the drivable path and the path guide points of the vehicle in each space piece are obtained based on a random uniform sampling and augmentation type search algorithm, and the heuristic search algorithm is fused to optimize the path between every two adjacent path guide points, so that the length of the driving path can be shortened, and the solving speed of the algorithm is improved; the method comprises the steps of connecting the drivable paths of all space slices to obtain a vehicle motion path, optimizing the path in the drivable area under the vehicle motion constraint by establishing a target optimization function and a constraint condition, finishing unmanned parking path planning, dynamically adjusting the vehicle safety distance aiming at vehicles with different task functions, improving the trafficability of the vehicles in a complex and variable environment, and realizing real-time parking of the unmanned vehicles.

Description

Unmanned parking path planning method for special road scene

Technical Field

The invention relates to the technical field of unmanned driving, in particular to an unmanned parking path planning method for a special road scene.

Background

The special road scene generally refers to a road with a low structuralization degree, and under the road scene, no lane lines and clear road boundaries exist, and the road is curved and complex, such as fields, mining areas, ports and the like. The unmanned parking path planning under the special road scene mainly means that an unmanned vehicle generates a safe drivable path inside the special road scene according to external environment information, the state of the vehicle and parking task requirements. The unmanned vehicle needs to generate a drivable safe parking path under the constraint of vehicle kinematics according to the requirements of parking tasks.

For the urban road environment, because the road is continuous and the road surface is flat, unmanned path planning can be easily realized. For a complicated and severe special road scene, because the scene is complicated and diversified and is mostly a large-scale scene, how to perform unmanned parking path real-time planning in the special road scene is one of the technical difficulties.

In recent years, a path planning method for automatically driving vehicles is provided, and parking path planning is completed through heuristic search, but the method mainly aims at small-range areas of urban and rural scenes, and does not relate to geographic large-range parking scenes and road complex curved scenes, and the heuristic search under the scenes can cause low completeness of a planned path, slow solving speed of an algorithm and cannot ensure real-time planning of the parking path.

Disclosure of Invention

In view of the above, the invention provides an unmanned parking path planning method for a special road scene, which is used for solving the problem of path planning inside the road scene with a low structuralization degree and realizing the rapid generation of a vehicle parking path meeting the vehicle kinematics constraint in a large-scale area.

The invention provides a method for planning an unmanned parking path in a special road scene, which comprises the following steps:

s1: obtaining a starting point P from a vehicle₀To the parking end point P_NWarp stop motion pointP₁,P₂,…,P_N-1(ii) a Wherein N is a positive integer;

s2: starting the vehicle from the starting point P by taking each passing stop point as a demarcation point₀To the parking end point P_NIs divided into N space slices S₁,S₂,…,S_N；

S3: judging the current space slice S_kTwo end points P of_k-1And P_kIs greater than a first threshold, k is 1,2, …, N; if yes, the current space slice S is processed_kThe interior position points are randomly and uniformly sampled, the sampling points are random path points, each random path point is connected with 3 other closest random path points, and the current space slice S is obtained by utilizing an extended search algorithm_kPossible travel path of, current space piece S_kRandom path points on the travelable path and the current space segment S_kTwo end points P of_k-1And P_kFor the current space slice S_kThe route guidance point of (1); if not, the current space slice S is divided into a plurality of space slices_kTwo end points P of_k-1And P_kConnected as a current space slice S_kPossible travel path of, current space piece S_kTwo end points P of_k-1And P_kFor the current space slice S_kThe route guidance point of (1);

s4: repeatedly executing the step S3, acquiring the travelable path and the path guidance point of the next space slice k being k +1 until the acquisition of the travelable paths and the path guidance points of all the space slices is completed;

s5: optimizing the path between every two adjacent path guide points by using a heuristic search algorithm;

s6: connecting the feasible driving paths of all the optimized space pieces to obtain a vehicle moving path;

s7: and establishing a target optimization function, vehicle kinematics constraint, path curvature constraint and peripheral obstacle constraint, and performing iterative optimization on the vehicle motion path by taking a path guide point on the vehicle motion path as an initial solution to obtain an optimal parking path.

In one possible implementation, in the inventionIn the provided unmanned parking path planning method for the special road scene, step S1 is to obtain the starting point P of the vehicle₀To the parking end point P_NWarp stop point P of₁,P₂,…,P_N-1The method specifically comprises the following steps:

receiving a parking terminal point P issued by a cloud intelligent platform through a global task receiving system in a vehicle-mounted computing unit_NAnd a series of stop points P₁,P₂,…,P_N-1Recording and storing; if the parking point does not exist, only the parking end point is recorded.

In a possible implementation manner, in the above unmanned parking path planning method for a special road scene provided by the present invention, in step S3, the current space slice S is obtained by using an extended search algorithm_kThe travelable path of (2) specifically includes:

the difference between the cost values of any two connected random path points is the space distance of the two random path points, and the current space slice S is selected_kEnd point P of_k-1To the current space slice S_kEnd point P of_kThe path with the smallest cost value is the current space slice S_kIs determined.

In a possible implementation manner, in the above unmanned parking path planning method for a special road scene provided by the present invention, step S5 is to optimize a path between every two adjacent path guidance points by using a heuristic search algorithm, which specifically includes:

taking any two adjacent path guide points as a starting point and an end point of a heuristic search algorithm, and according to the distance d between the starting point and the end point_objAdjusting the extended step length of the heuristic search algorithm to

And according to the distance d between the extended point and the end point of the heuristic search algorithm_goalAdjusting the weight w of the heuristic function_hIf, if

Then w_hIf 1, then

Then 1 < w_hLess than or equal to 1.2; based on the Ackerman bicycle model of the vehicle, path optimization between two adjacent path guide points is completed; and the course angle of the vehicle when reaching the terminal is consistent with the terminal course angle, and the terminal course angle is an included angle between the terminal and the next route guidance point of the terminal.

In a possible implementation manner, in the above unmanned parking path planning method for a special road scene provided by the present invention, step S7, establishing an objective optimization function, a vehicle kinematics constraint, a path curvature constraint, and a peripheral obstacle constraint, and iteratively optimizing a vehicle motion path with a path guide point on the vehicle motion path as an initial solution to obtain an optimal parking path specifically includes:

establishing a target function related to path curvature cost, obstacle cost and path length cost by taking the vehicle motion path as an initial solution; wherein the path curvature cost is the sum of the curvatures of all path guide points on the vehicle motion path; the obstacle cost is d_obsAnd a second threshold value d_thdDifference of (d)_obsDistance between route guide point and nearest obstacle, if d_obs≥d_thdThe obstacle cost is 0, if d_obs＜d_thdThen the cost of the obstacle is d corresponding to all the route guidance points_obsAnd d_thdThe sum of the squares of the differences; the path length cost is the sum of the distances between every two adjacent path guide points on the vehicle motion path; under the condition of meeting the vehicle kinematics constraint, the path curvature constraint and the peripheral obstacle constraint, the sum of the path curvature cost, the obstacle cost and the path length cost of the target optimization function is minimized, and the optimal parking path is obtained.

In a possible implementation manner, in the above unmanned parking path planning method for a special road scene provided by the present invention, after performing step S6, connecting the feasible driving paths of all optimized space pieces to obtain a vehicle motion path, performing step S7, establishing an objective optimization function, and vehicle kinematic constraint, path curvature constraint, and peripheral obstacle constraint, and performing iterative optimization on the vehicle motion path with a path guide point on the vehicle motion path as an initial solution to obtain an optimal parking path, the method further includes:

all path guide points on the vehicle motion path are used as control nodes of a B spline, the B spline is used for carrying out curve fitting on the vehicle motion path, twice of the vehicle length is used as an interval step length, path interpolation of the vehicle motion path is completed, and the interpolation nodes are used for replacing original path guide points on the vehicle motion path.

The unmanned parking path planning method for the special road scene is provided aiming at the problem of low planning efficiency of unmanned parking path planning in special road environments such as fields, mining areas and ports. The road space from the starting point to the parking end point of the vehicle is divided into a plurality of space pieces by taking the passing point as a demarcation point, so that the solving speed of the algorithm can be improved; the drivable path and the path guide points of the vehicle in each space piece are obtained based on a random uniform sampling and augmentation type search algorithm, and the heuristic search algorithm is fused to optimize the path between every two adjacent path guide points, so that the length of the driving path can be shortened, and the solving speed of the algorithm is improved; real-time parking of the unmanned vehicle is realized; the method comprises the steps of connecting the drivable paths of all space slices to obtain a vehicle motion path, optimizing the path in the drivable area under the constraint of vehicle motion by establishing a target optimization function and a constraint condition, finishing unmanned parking path planning, dynamically adjusting the vehicle safety distance aiming at vehicles with different task functions, reducing the curvature of the unmanned vehicle parking path, improving the trafficability of the vehicle in a complex and variable environment, and realizing real-time parking of the unmanned vehicle.

Drawings

FIG. 1 is a flow chart of a method for unmanned parking path planning for special road scenarios provided by the present invention;

fig. 2 is a schematic view of the special road scene space segment division and parking path planning in embodiment 1 of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only illustrative and are not intended to limit the present invention.

The invention provides a method for planning an unmanned parking path in a special road scene, which comprises the following steps as shown in figure 1:

s1: obtaining a starting point P from a vehicle₀To the parking end point P_NWarp stop point P of₁,P₂,…,P_N-1(ii) a Wherein N is a positive integer;

s2: starting the vehicle from the starting point P by taking each passing stop point as a demarcation point₀To the parking end point P_NIs divided into N space slices S₁,S₂,…,S_N；

S3: judging the current space slice S_kTwo end points P of_k-1And P_kIs greater than a first threshold value d_mK is 1,2, …, N; if yes, the current space slice S is processed_kThe interior position points are randomly and uniformly sampled, the sampling points are random path points, each random path point is connected with 3 other closest random path points, and the current space slice S is obtained by utilizing an extended search algorithm_kPossible travel path of, current space piece S_kRandom path points on the travelable path and the current space segment S_kTwo end points P of_k-1And P_kFor the current space slice S_kThe route guidance point of (1); if not, the current space slice S is divided into a plurality of space slices_kTwo end points P of_k-1And P_kConnected as a current space slice S_kOf the current space piece, two end points P of the current space piece_k-1And P_kFor the current space slice S_kThe route guidance point of (1);

s4: repeatedly executing the step S3, acquiring the travelable path and the path guidance point of the next space slice k being k +1 until the acquisition of the travelable paths and the path guidance points of all the space slices is completed;

s5: optimizing the path between every two adjacent path guide points by using a heuristic search algorithm;

s6: connecting the feasible driving paths of all the optimized space pieces to obtain a vehicle moving path;

s7: and establishing a target optimization function, vehicle kinematics constraint, path curvature constraint and peripheral obstacle constraint, and performing iterative optimization on the vehicle motion path by taking a path guide point on the vehicle motion path as an initial solution to obtain an optimal parking path.

The following describes a specific implementation of the above-mentioned unmanned parking path planning method for a special road scene according to a specific embodiment.

Example 1:

the first step is as follows: obtaining a starting point P from a vehicle₀To the parking end point P_NWarp stop point P of₁,P₂,…,P_N-1(ii) a Wherein N is a positive integer.

Each vehicle can be provided with a vehicle-mounted computing unit for running a path planning algorithm, and a global task receiving system in the vehicle-mounted computing unit receives a parking terminal point P issued by a cloud intelligent platform_NAnd a series of stop points P₁,P₂,…,P_N-1The vehicle-mounted computing unit is used for parking the vehicle at the end point P_NAnd all the passing stop points P₁,P₂,…,P_N-1And recording and storing. If the parking point does not exist, only the parking end point is recorded. As shown in fig. 2, from the vehicle starting point P₀To the parking end point P₅With four warp stop points P in between₁、P₂、P₃And P₄Dotted line A₁And dotted line A₂Representing the left and right boundaries of the road, respectively.

The second step is that: starting the vehicle from the starting point P by taking each passing stop point as a demarcation point₀To the parking end point P_NIs divided into N space slices S₁,S₂,…,S_N。

From the vehicle starting point P₀To the first passing stop point P₁Is the first space slice S₁From the firstWarp stop point P₁To a second point of warp stop P₂In the area of the second space piece S₂From the second point of transit P₂To a third point of warp stop P₃In the third space segment S₃By analogy, from the N-1 th warp stop point P_N-1To the parking end point P_NIn the Nth space slice S_N. As shown in FIG. 2, the vehicle is started from point P₀To the parking end point P₅Is divided into 5 space slices S₁、S₂、S₃、S₄、S₅。

The third step: judging the current space slice S_kTwo end points P of_k-1And P_kIs greater than a first threshold, k is 1,2, …, N; if yes, the current space slice S is processed_kThe interior position points are randomly and uniformly sampled, the sampling points are random path points, each random path point is connected with 3 other closest random path points, and the current space slice S is obtained by utilizing an extended search algorithm_kPossible travel path of, current space piece S_kRandom path points on the travelable path and the current space segment S_kTwo end points P of_k-1And P_kFor the current space slice S_kThe route guidance point of (1); if not, the current space slice S is divided into a plurality of space slices_kTwo end points P of_k-1And P_kConnected as a current space slice S_kPossible travel path of, current space piece S_kTwo end points P of_k-1And P_kFor the current space slice S_kThe path guidance point of (1).

In particular, from the first space piece S₁Starting to judge the first space slice S₁Two end points of (2), i.e. the vehicle starting point P₀And a first warp stop point P₁Whether the distance therebetween is greater than a first threshold; if yes, utilizing the vehicle-mounted computing unit to carry out calculation on the first space slice S₁The interior position points are randomly and uniformly sampled, the sampling points are random path points, each random path point is connected with 3 other random path points which are nearest, and a first space slice S is obtained by utilizing an augmentation search algorithm₁The first space piece S₁Random route points on the travelable route and the first space segment S₁Two end points P of₀And P₁For the first space slice S₁The route guidance point of (1); if not, the first space slice S₁Two end points of (2), i.e. the vehicle starting point P₀And a first warp stop point P₁Connected as a first space piece S₁The first space piece S₁Two end points of (2), i.e. the vehicle starting point P₀And a first warp stop point P₁For the first space slice S₁The path guidance point of (1). Then, for the second space slice S₂Making a judgment to obtain a second space slice S₂The possible driving path and the path guiding point, the concrete process and the first space slice S₁Similarly, the description is omitted here. And by analogy, judging all the space pieces so as to obtain the travelable paths and path guide points of all the space pieces. As shown in fig. 2, from the vehicle starting point P₀To the first passing stop point P₁First space slice S₁With 3 path guidance points Q₀、Q₁And Q₂Wherein Q is₀Is a vehicle starting point P₀，Q₁Being random path points, Q₂Is an end point P₁(ii) a From the first point of warp stop P₁To a second point of warp stop P₂Second space piece S₂With 2 path guidance points Q₂And Q₃Wherein Q is₃Are all endpoints P₂(ii) a From the second point of warp stop P₂To a third point of warp stop P₃Third space piece S₃With 2 path guidance points Q₃And Q₄Wherein Q is₄Is an end point P₃(ii) a From a third point of warp stop P₃To a fourth point of warp stop P₄Fourth space piece S₄With 3 path guidance points Q₄、Q₅And Q₆Wherein Q is₅Being random path points, Q₆Is an end point P₄(ii) a From the fourth point of stoppage P₄To the parking end point P₅The fifth space piece S₅With 4 path guidance points Q₆,Q₇，Q₈And Q₉Wherein Q is₇And Q₈Are all random path points, Q₉For parking terminal P₅。

In the current space slice S_kMaking a judgment, and the current space slice S_kTwo end points P of_k-1And P_kWhen the distance between the space slices is greater than the first threshold value, the current space slice S needs to be processed_kRandom uniform sampling is carried out to obtain random path points, each random path point is connected with 3 other random path points which are nearest, and therefore an end point P from the current space slice is obtained_k-1To the current space slice S_kEnd point P of_kHaving many paths, then, using an augmented search algorithm, the current space slice S can be obtained from many paths_kOf the current space piece S, in particular_kThe travelable path of (2) can be obtained by: the difference between the cost values of any two connected random path points is the spatial distance of the two random path points, for example, a path is randomly selected from a plurality of paths, and the starting point of the path, i.e. the current space slice S, is assumed_kEnd point P of_k-1Has a cost value of a, the starting point P of the path_k-1The spatial distance between the random path point and the connected random path point is b, and the random path point and the starting point P are_k-1The cost value of the connected random path point is a + b; traversing an endpoint P from a current spatial slice_k-1To the current space slice S_kEnd point P of_kAfter all the paths are selected, the current space slice S is selected_kEnd point P of_k-1To the current space slice S_kEnd point P of_kAs the current space slice S, the path with the smallest cost value_kIs determined.

The fourth step: after the travelable paths and the path guide points of all the space pieces are obtained, optimizing the path between every two adjacent path guide points by using a heuristic search algorithm.

Specifically, any two adjacent path guide points are taken as a starting point and an end point of a heuristic search algorithm, and the distance d between the starting point and the end point is determined according to the distance d between the starting point and the end point_objAdjusting the extended step length of the heuristic search algorithm to

And according to the distance d between the extended point and the end point of the heuristic search algorithm_goalAdjusting the weight w of the heuristic function_hIf, if

Then w_hIf 1, then

Then 1 < w_hLess than or equal to 1.2; then, based on the Ackerman bicycle model of the vehicle, path optimization between the two adjacent path guide points is completed; it should be noted that the heading angle of the vehicle when reaching the end point needs to be consistent with the end point heading angle, and the end point heading angle is an included angle between the end point and the next route guidance point of the end point.

The fifth step: after optimizing the paths between all the adjacent two path guide points in all the space pieces, connecting the feasible driving paths of all the optimized space pieces to obtain the vehicle movement path.

And a sixth step: all path guide points on the vehicle motion path are used as control nodes of a B spline, the B spline is used for carrying out curve fitting on the vehicle motion path, twice of the vehicle length is used as an interval step length, path interpolation of the vehicle motion path is completed, and the interpolation nodes are used for replacing original path guide points on the vehicle motion path.

The seventh step: and establishing a target optimization function, vehicle kinematics constraint, path curvature constraint and peripheral obstacle constraint, and performing iterative optimization on the vehicle motion path by taking a path guide point on the vehicle motion path subjected to path interpolation as an initial solution to obtain an optimal parking path.

Establishing a target function related to path curvature cost, obstacle cost and path length cost by taking a vehicle motion path as an initial solution; the path curvature cost is the sum of the curvatures of all path guide points on the vehicle motion path; the cost of the obstacle is d_obsAnd a second threshold value d_thdDifference of (d)_obsDistance between guide point of path and nearest obstacleIf d is away from_obs≥d_thdThe obstacle cost is 0, if d_obs＜d_thdThen the cost of the obstacle is d corresponding to all the route guidance points_obsAnd d_thdThe sum of the squares of the differences; the path length cost is the sum of the distances between every two adjacent path guide points on the vehicle motion path; under the condition of meeting the vehicle kinematics constraint, the path curvature constraint and the peripheral obstacle constraint, the sum of the path curvature cost, the obstacle cost and the path length cost of the target optimization function is minimized, and therefore the optimal parking path is obtained.

The unmanned parking path planning method for the special road scene is provided aiming at the problem of low planning efficiency of unmanned parking path planning in special road environments such as fields, mining areas and ports. The road space from the starting point to the parking end point of the vehicle is divided into a plurality of space pieces by taking the passing point as a demarcation point, so that the solving speed of the algorithm can be improved; the drivable path and the path guide points of the vehicle in each space piece are obtained based on a random uniform sampling and augmentation type search algorithm, and the heuristic search algorithm is fused to optimize the path between every two adjacent path guide points, so that the length of the driving path can be shortened, the solving speed of the algorithm is improved, and the real-time parking of the unmanned vehicle is realized; the method comprises the steps of connecting the drivable paths of all space slices to obtain a vehicle motion path, optimizing the path in the drivable area under the constraint of vehicle motion by establishing a target optimization function and a constraint condition, finishing unmanned parking path planning, dynamically adjusting the vehicle safety distance aiming at vehicles with different task functions, reducing the curvature of the unmanned vehicle parking path, improving the trafficability of the vehicle in a complex and variable environment, and realizing real-time parking of the unmanned vehicle.

It will be apparent to those skilled in the art that various changes and modifications may be made in the present invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.

Claims (6)

1. A method for planning an unmanned parking path for a special road scene is characterized by comprising the following steps:

s1: obtaining a starting point P from a vehicle₀To the parking end point P_NWarp stop point P of₁,P₂,…,P_N-1(ii) a Wherein N is a positive integer;

s2: starting the vehicle from the starting point P by taking each passing stop point as a demarcation point₀To the parking end point P_NIs divided into N space slices S₁,S₂,…,S_N；

S3: judging the current space slice S_kTwo end points P of_k-1And P_kIs greater than a first threshold, k is 1,2, …, N; if yes, the current space slice S is processed_kThe interior position points are randomly and uniformly sampled, the sampling points are random path points, each random path point is connected with 3 other closest random path points, and the current space slice S is obtained by utilizing an extended search algorithm_kPossible travel path of, current space piece S_kRandom path points on the travelable path and the current space segment S_kTwo end points P of_k-1And P_kFor the current space slice S_kThe route guidance point of (1); if not, the current space slice S is divided into a plurality of space slices_kTwo end points P of_k-1And P_kConnected as a current space slice S_kPossible travel path of, current space piece S_kTwo end points P of_k-1And P_kFor the current space slice S_kThe route guidance point of (1);

s4: repeatedly executing the step S3, acquiring the travelable path and the path guidance point of the next space slice k being k +1 until the acquisition of the travelable paths and the path guidance points of all the space slices is completed;

s5: optimizing the path between every two adjacent path guide points by using a heuristic search algorithm;

s6: connecting the feasible driving paths of all the optimized space pieces to obtain a vehicle moving path;

s7: and establishing a target optimization function, vehicle kinematics constraint, path curvature constraint and peripheral obstacle constraint, and performing iterative optimization on the vehicle motion path by taking a path guide point on the vehicle motion path as an initial solution to obtain an optimal parking path.

2. The method for unmanned vehicle parking path planning for special road scenes as claimed in claim 1, wherein step S1 obtains the point P from the vehicle start point₀To the parking end point P_NWarp stop point P of₁,P₂,…,P_N-1The method specifically comprises the following steps:

receiving a parking terminal point P issued by a cloud intelligent platform through a global task receiving system in a vehicle-mounted computing unit_NAnd a series of stop points P₁,P₂,…,P_N-1Recording and storing; if the parking point does not exist, only the parking end point is recorded.

3. The method for unmanned vehicle parking path planning for special road scenes as claimed in claim 1, wherein in step S3, the current space piece S is obtained by using an extended search algorithm_kThe travelable path of (2) specifically includes:

the difference between the cost values of any two connected random path points is the space distance of the two random path points, and the current space slice S is selected_kEnd point P of_k-1To the current space slice S_kEnd point P of_kThe path with the smallest cost value is the current space slice S_kIs determined.

4. The method for planning the unmanned parking path for the special road scene as claimed in claim 1, wherein step S5, using a heuristic search algorithm, optimizes the path between each two adjacent path guidance points, specifically comprising:

taking any two adjacent path guide points as a starting point and an end point of a heuristic search algorithm, and according to the distance d between the starting point and the end point_objAdjusting the extended step length of the heuristic search algorithm to

And according to the distance d between the extended point and the end point of the heuristic search algorithm_goalAdjusting the weight w of the heuristic function_hIf, if

Then w_hIf 1, then

Then 1 < w_hLess than or equal to 1.2; based on the Ackerman bicycle model of the vehicle, path optimization between two adjacent path guide points is completed; and the course angle of the vehicle when reaching the terminal is consistent with the terminal course angle, and the terminal course angle is an included angle between the terminal and the next route guidance point of the terminal.

5. The unmanned parking path planning method for the special road scene as claimed in claim 1, wherein step S7, establishing an objective optimization function, a vehicle kinematics constraint, a path curvature constraint and a peripheral obstacle constraint, and iteratively optimizing the vehicle motion path with a path guide point on the vehicle motion path as an initial solution to obtain an optimal parking path specifically comprises:

establishing a target function related to path curvature cost, obstacle cost and path length cost by taking the vehicle motion path as an initial solution; wherein the path curvature cost is the sum of the curvatures of all path guide points on the vehicle motion path; the obstacle cost is d_obsAnd a second threshold value d_thdDifference of (d)_obsDistance between route guide point and nearest obstacle, if d_obs≥d_thdThe obstacle cost is 0, if d_obs＜d_thdThen the cost of the obstacle is d corresponding to all the route guidance points_obsAnd d_thdThe sum of the squares of the differences; the path length cost is the sum of the distances between every two adjacent path guide points on the vehicle motion path; under the condition of satisfying the vehicle kinematic constraint, the path curvature constraint and the peripheral obstacle constraintAnd the sum of the path curvature cost, the obstacle cost and the path length cost of the target optimization function is minimum, so that the optimal parking path is obtained.

6. The unmanned parking path planning method for special road scenes as claimed in any one of claims 1 to 5, wherein after executing step S6, connecting the feasible driving paths of all optimized space slices to obtain a vehicle motion path, executing step S7, establishing an objective optimization function, and vehicle kinematics constraint, path curvature constraint and peripheral obstacle constraint, and performing iterative optimization on the vehicle motion path with a path guide point on the vehicle motion path as an initial solution to obtain an optimal parking path, further comprises:

all path guide points on the vehicle motion path are used as control nodes of a B spline, the B spline is used for carrying out curve fitting on the vehicle motion path, twice of the vehicle length is used as an interval step length, path interpolation of the vehicle motion path is completed, and the interpolation nodes are used for replacing original path guide points on the vehicle motion path.

Images