A Parking Trajectory Planning Method and Device for a Four-Wheel Independent Steering Vehicle

Abstract

The present invention discloses a parking trajectory planning method and device for a four-wheel independent steering vehicle, which includes: S1, obtaining the initial position of the vehicle, the target parking position, and the grid map of the parking scenario, using the target parking position as the search starting point, the initial position of the vehicle as the end point, and performing global path planning with an improved hybrid A* algorithm to obtain the original global path; the improved hybrid A* algorithm includes a child node expansion step, and this child node expansion step adopts a child node expansion method corresponding to the vehicle motion mode set in advance; S2, using the mode switching nodes on the original global path as endpoints, segmenting the original global path, performing path optimization and speed planning on each segment of the path to obtain the optimized global trajectory. The present invention can give full play to the flexibility of the four-wheel independent steering vehicle.

Description

Technical Field

[0001] The present invention relates to the technical field of autonomous driving, and particularly to a parking trajectory planning method and device for a four-wheel independent steering vehicle. Background Art

[0002] For a four-wheel independent steering vehicle, all four wheels can be independently controlled, and it has three motion modes: front and rear wheels with opposite directions, rotation in place, and front and rear wheels with the same direction. It is a vehicle type with high flexibility and high passability. When facing some narrow parking environments, compared with a front-wheel steering vehicle, due to its multiple motion modes, a four-wheel independent steering vehicle can better cope with narrow environments. In recent years, automobile manufacturers and universities in various countries have vigorously invested manpower and financial resources in the development and research of four-wheel independent drive vehicles. With the development of autonomous driving technology, due to its excellent motion ability, four-wheel independent steering vehicles will surely be widely used in the field of autonomous driving.

[0003] The existing research on the trajectory planning of four-wheel independent steering vehicles basically follows the trajectory planning methods of front-wheel steering vehicles, which mainly include methods based on graph search, sampling, and optimization. However, these trajectory planning methods for front-wheel steering vehicles do not fully consider the in-place rotation and front and rear wheel same-direction motion modes of four-wheel independent steering vehicles, and cannot give full play to the high flexibility of the motion of four-wheel independent steering vehicles. Related research also includes path planning of omnidirectional mobile robots, etc., but these studies only plan the vehicle path through a single motion mode. To sum up, the existing research methods are not applicable to four-wheel independent steering vehicles. Summary of the Invention

[0004] The purpose of the present invention is to provide a parking trajectory planning method and device for a four-wheel independent steering vehicle, which can make full use of the three motion modes of front and rear wheel opposite directions, rotation in place, and front and rear wheel same direction of the four-wheel independent steering vehicle, and give full play to its high flexibility characteristics.

[0005] To achieve the above object, the present invention provides a parking trajectory planning method for a four-wheel independent steering vehicle, which includes:

[0006] S1, obtaining the initial position of the vehicle, the target parking position, and the grid map of the parking scene, using the target parking position as the search starting point and the initial position of the vehicle as the end point, and performing global path planning with an improved hybrid A* algorithm to obtain an original global path; wherein, the improved hybrid A* algorithm includes a sub-node expansion step, and this sub-node expansion step adopts a sub-node expansion method corresponding to the vehicle motion mode set in advance;

[0007] S2, taking the mode switching nodes on the original global path as endpoints, segmenting the original global path, and performing path optimization and speed planning on each segment of the path to obtain an optimized global trajectory.

[0008] Further, in step S1, the child node expansion method corresponding to the front and rear wheel reverse movement mode in the vehicle movement mode generates (k1 + 2) child nodes. The pose description of the left-rotating child node is expressed by the following formula (1), the pose description of the right-rotating child node is expressed by the following formula (2), and the pose description of the straight-going child node is expressed by the following formula (3):

[0009]

[0010] In the formula, (x1, y1) is the coordinate of one of the (k1 + 2) child nodes in the earth coordinate system, (x, y) is the coordinate of the parent node in the earth coordinate system, R is the turning radius, and i is all integers from 1 to k1 / 4, α is the vehicle heading angle at the parent node, Δα is the yaw angle of the vehicle, l1 is the step size, and α1 is the vehicle heading angle at one of the (k1 + 2) child nodes.

[0011] Further, in step S1, the child node expansion method corresponding to the in-situ rotation movement mode in the vehicle movement mode generates k2 child nodes described by the following formula (5):

[0012]

[0013] In the formula, (x2, y2) is the coordinate of one of the k2 child nodes in the earth coordinate system, (x, y) is the coordinate of the parent node in the earth coordinate system, α is the vehicle heading angle at the parent node, i is all integers from 1 to k2, and α2 is the heading angle of one of the k2 child nodes.

[0014] Further, in step S1, the child node expansion method corresponding to the front and rear wheel forward movement mode in the vehicle movement mode generates k3 child nodes described by the following formula (7):

[0015]

[0016] In the formula, (x3, y3) is the coordinate of one of the k3 child nodes in the earth coordinate system, (x, y) is the coordinate of the parent node in the earth coordinate system, α is the vehicle heading angle at the parent node, l2 is the step size, α3 is the heading angle of one of the k3 child nodes, and the front wheel steering angle j is all integers from 1 to k3 / 4.

[0017] Further, in the process of the "improved hybrid A* algorithm" in step S1, the expansion mode information of each generated node is recorded. When calculating the cost value of a child node, when the expansion mode information of the parent node is the same as that of the child node, the total cost value of the child node is expressed as f(n) = g(n) + h(n); when the expansion mode information of the parent node is different from that of the child node, the total cost value of the child node is expressed as f(n) = g(n) + h(n) + p(n), where f(n), g(n), h(n), and p(n) respectively represent the total cost value, path cost value, heuristic cost value, and mode switching cost value of node n.

[0018] Further, when calculating the heuristic cost value g(n) of a child node in the "improved hybrid A* algorithm" in step S1, when the expansion mode information of the child node is the front-rear wheel reverse, in-situ rotation, and front-rear wheel forward movement modes respectively, the heuristic cost values g(n) are described as g1(n), g2(n), and g3(n) provided by the following formulas:

[0019] g1(n) = l1·(ω l1 + ω b1 + ω t1 ·Δα) (4)

[0020] g2(n) = ω r ·|α2 - α| (6)

[0021] g3(n) = l2·(ω l2 + ω b2 + ω t2 ·β) (8)

[0022] In the formulas, ω l1 、ω b1 and ω t1 are respectively the path length penalty coefficient, reverse penalty coefficient, and turning penalty coefficient corresponding to the expansion mode of the front-rear wheel reverse movement mode; ω r is the rotation angle penalty coefficient corresponding to the expansion mode of the in-situ rotation mode; ω l2 、ω b2 and ω t1 are respectively the path length penalty coefficient, reverse penalty coefficient, and wheel turning angle penalty coefficient corresponding to the expansion mode of the front-rear wheel forward movement mode.

[0023] Further, when calculating the mode switching cost value p(n) of a child node in the "improved hybrid A* algorithm" in step S1, if it is the mode switching between the front-rear wheel reverse and in-situ rotation modes, p(n) is described as the constant p1; if it is the mode switching between the front-rear wheel reverse and front-rear wheel forward modes, p(n) is described as the constant p2; if it is the mode switching between the front-rear wheel forward and in-situ rotation modes, p(n) is described as the constant p3.

[0024] Further, in step S2, the original global path is segmented into front and rear wheel anisotropic motion path segments and front and rear wheel unidirectional motion path segments, and the front and rear wheel anisotropic motion path segments are optimized using a gradient-based numerical optimization method.

[0025] The present invention provides a parking trajectory planning device for a four-wheel independent steering vehicle, comprising:

[0026] An environmental perception module, which is used to generate parking space information, grid maps and vehicle initial position information;

[0027] The parking space management module is used to receive the parking space information, grid map and vehicle initial position information transmitted by the environment perception module, select an optimal parking space for the vehicle, and then generate target parking position information according to the selected parking space;

[0028] a parking trajectory planning module, which is used to receive the vehicle initial position information and grid map transmitted by the environment perception module and the target parking position information transmitted by the parking space management module, and use the optimized global trajectory generated by the parking trajectory planning method for a four-wheel independent steering vehicle according to any one of claims 1 to 8 as the parking trajectory;

[0029] The vehicle control module is used to complete the parking task according to the trajectory information after receiving the parking trajectory information transmitted by the parking trajectory planning module. The vehicle control module has a control function and is used to control the vehicle to complete the parking task. After receiving the parking trajectory information transmitted by the parking trajectory planning module, the vehicle control module controls the vehicle's steering wheel, accelerator, brake and other actuators according to the trajectory information to complete the parking task.

[0030] Front-wheel steering vehicles tend to plan more reversing and turning trajectories when parking in some narrow roads, and may even fail to plan. Compared with front-wheel steering vehicles, four-wheel independent steering vehicles can more flexibly complete parking trajectory planning in narrow roads and narrow parking spaces. In order to give full play to the excellent movement ability of four-wheel independent steering vehicles in parking trajectory planning, the present invention designs a parking trajectory planning method and system for four-wheel independent steering vehicles by combining the three movement modes of four-wheel independent steering vehicles.

[0031] The present invention improves the hybrid A* algorithm and designs three node expansion methods for the three motion modes of four-wheel independent steering vehicles. It effectively combines the three motion modes of the front and rear wheels of the four-wheel independent steering vehicle, rotating in situ, and the front and rear wheels in the same direction into the graph search algorithm, maximizing the flexibility of the trajectory planning of the four-wheel steering vehicle, so that the vehicle can flexibly switch the motion mode to cope with various environments. Especially for some parking scenarios in narrow spaces, the present invention greatly increases the trajectory quality and success rate of vehicle trajectory planning. Description of the Drawings

[0032] Figure 1 This is a flowchart of parking trajectory planning in an embodiment of the present invention.

[0033] Figure 2 This is a schematic diagram of three sub - node expansion methods in an embodiment of the invention: (a) is the sub - node expansion method for the forward and rear wheels moving in opposite directions; (b) is the sub - node expansion method for rotating in place; (c) is the sub - node expansion method for the forward and rear wheels moving in the same direction.

[0034] Figure 3 This is a flowchart of the improved hybrid A* algorithm in an embodiment of the invention.

[0035] Figure 4 This is a schematic diagram of the global path generated by combining the forward and rear wheels moving in opposite directions and the forward and rear wheels moving in the same direction in an embodiment of the invention.

[0036] Figure 5 This is a schematic diagram of the global path generated by combining the forward and rear wheels moving in opposite directions and rotating in place in an embodiment of the invention.

[0037] Figure 6 This is a schematic diagram of the structure of the parking trajectory planning system for a four - wheel independently - steered vehicle in an embodiment of the invention. Detailed Embodiment

[0038] The present invention will be described in detail below with reference to the drawings and embodiments.

[0039] As Figure 1 shown, the parking trajectory planning method for a four - wheel independently - steered vehicle provided by the embodiment of the present invention includes:

[0040] S1. Obtain the initial position of the vehicle, the target parking position, and the grid map of the parking scenario. Use the target parking position as the search starting point n start , and the initial position of the vehicle as the end point. Perform global path planning using the improved hybrid A* algorithm to obtain the original global path. Among them, in this embodiment, the search starting point and the search target point are interchanged, and the path search is performed in the reverse direction to obtain the parking - in path. Since the target parking position of the vehicle is usually narrower than the initial position of the vehicle, searching for the parking - out path to reversely generate the parking path can find the path end point faster, which can improve the search efficiency of the algorithm.

[0041] S2. Take the mode - switching nodes on the original global path as endpoints, segment the original global path, and perform path optimization and speed planning on each segment of the path to obtain the optimized global trajectory.

[0042] In one embodiment, as Figure 3 shown, the improved hybrid A* algorithm in step S1 specifically includes:

[0043] S11, establish an open list and a closed list, initialize both the open list and the closed list to empty lists, and set the search starting point n start Put in the open list.

[0044] S12, check whether there is a node in the open list, if it is determined to be yes, go to step 13, if it is determined to be no, the path planning fails.

[0045] S13, open the child node n with the smallest total cost value in the list i Pop it from the open list and put it into the closed list. The total cost can be described as f(n)=g(n)+h(n), where f(n), g(n), and h(n) represent the total cost, path cost, and heuristic cost of node n, respectively.

[0046] S14, determine the node n in the above step S13 i Can it be expanded to target node n? goal Or generate the target node n goal If the judgment is yes, the path planning ends and the original global path is successfully searched; if the judgment is no, the process goes to step S15.

[0047] S15, for the node n in the above step S13 i , through the node expansion method, we get several child nodes ni +1 First, perform collision detection on the child nodes. The child nodes without collision in the detection results are feasible nodes. The child nodes with collision in the detection results are eliminated, and then the posture information and path cost value g(n) of all feasible child nodes are calculated. i+1 ), heuristic cost h(n i+1 ), mode switching cost p(n i+1 ), total generation value f(n i+1 ), as well as recording node expansion mode, wheel angle information, vehicle control direction information (forward / reverse), parent node information and other node information.

[0048] S16, according to whether the feasible child node in step S15 is in the open list or the closed list, record the parent node information on the optimal path to the feasible child node.

[0049] S17, looping step S12 to step S16 until the condition for ending the path planning is met, then the algorithm ends.

[0050] In one embodiment, the improved hybrid A* algorithm includes a subnode expansion step, which adopts a pre-set subnode expansion method corresponding to the vehicle motion mode. Figure 2As shown, the vehicle motion modes are: front and rear wheels moving in opposite directions motion mode, in-situ rotation motion mode, and front and rear wheels moving in the same direction motion mode. To flexibly utilize these three motion modes, the sub-node expansion methods in the "improved hybrid A* algorithm" in step S1 are different.

[0051] For example, in one embodiment, as Figure 4 and Figure 5 shown, in the case where the vehicle is in the front and rear wheels moving in opposite directions motion mode, the sub-node expansion method in the "improved hybrid A* algorithm" in step S1 generates (k1 + 2) sub-nodes by considering vehicle kinematic constraints and expanding node position information and angle information simultaneously. k1 is an integer multiple of 4. The pose description of the left rotation sub-node is given by the following formula (1), the pose description of the right rotation sub-node is given by the following formula (2), and the pose description of the straight-line sub-node is given by the following formula (3):

[0052]

[0053] In the above formula, (x, y, α) is the pose of the parent node in the earth coordinate system, (x, y) is the coordinate of the parent node in the earth coordinate system, α is the vehicle heading angle at the parent node, and (x1, y1) is the coordinate of one of the (k1 + 2) sub-nodes in the earth coordinate system, α1 is the vehicle heading angle at one of the (k1 + 2) sub-nodes, l1 is the step size, R is the turning radius, and i is all integers from 1 to k1 / 4, L is the wheelbase, and θ is the maximum wheel steering angle.

[0054] In one embodiment, in the case where the vehicle is in the front and rear wheels moving in opposite directions motion mode, the path cost value g(n) is described as g1(n) provided by the following formula (4):

[0055] g1(n) = l1·(ω l1 + ω b1 + ω t1 ·Δα) (4)

[0056] In the formula, ω l1 、ω b1 and ω t1 are respectively the path length penalty coefficient, reverse penalty coefficient, and turning penalty coefficient in the front and rear wheels moving in opposite directions motion mode. It should be noted that ω l1 is generally set to 1, and then based on this, small penalties are imposed on reverse and turning. For example, ω b1 is 0.5, and ω t1 is 0.2.

[0057] Of course, in one embodiment, during the process of "improved hybrid A* algorithm" in step S1, when the parent node expands a child node with a motion pattern inconsistent with its own, it indicates that a motion pattern switch occurs at this node. Then the total cost value is expressed as f(n) = g(n) + h(n) + p(n), where f(n), g(n), h(n), and p(n) represent the total cost value, path cost value, heuristic cost value, and pattern switch cost value of node n, respectively. Moreover, the specific value of the pattern switch cost value p(n) is related to the motion pattern of the parent node.

[0058] Then, in the case where the vehicle is in a front-rear wheel reverse motion pattern, the pattern switch cost value p(n) is set to p1. To prevent unnecessary pattern switches, and since the front-rear wheel reverse motion pattern includes in-place rotation and requires stopping, it is relatively large. Therefore, the specific value of p1 can be set within the range of 2 to 3.

[0059] Also, for example, in one embodiment, as Figure 5 shown, in the case where the vehicle is in an in-place rotation pattern, the child node expansion method in the "improved hybrid A* algorithm" in step S1 only expands the angular information of the node and does not expand the position information of the node, generating k2 child nodes, where k2 is an integer multiple of 4. The pose description of the child nodes is given by the following formula (5):

[0060]

[0061] In the formula, (x2, y2) are the coordinates of one of the k2 child nodes in the earth coordinate system, (x, y) are the coordinates of the parent node in the earth coordinate system, α is the current heading angle of the vehicle, i is all integers from 1 to k2, and α2 is the heading angle of one of the k2 child nodes.

[0062] In one embodiment, in the case where the vehicle is in an in-place rotation pattern, the path cost value g(n) is described by g2(n) provided in the following formula (6):

[0063] g2(n) = ω r ·|α2 - α| (6)

[0064] In the formula, ω r is the rotation angle penalty coefficient in the in-place rotation pattern, and the specific value can be set to 0.5.

[0065] In one embodiment, in the case where the vehicle is in an in-place rotation pattern, the pattern switch cost value p(n) is described as p2. Since the pattern switch of p2 can be performed at low speed, it can be set to be relatively small. For example, the specific value of p2 is set to 1.

[0066] Another example: In one embodiment, as Figure 4As shown, in the case where the vehicle is in the front and rear wheel same-direction movement mode, the child node expansion method in the "improved hybrid A* algorithm" in step S1 only expands the vehicle position information and does not expand the angle information, generating k3 child nodes, where k3 is an integer multiple of 4. The pose description of the child node is given by the following formula (7):

[0067]

[0068] In the above formula, l2 is the step size, and β is the wheel rotation angle. j is all integers from 1 to k3 / 4.

[0069] In one embodiment, in the case where the vehicle is in the front and rear wheel same-direction movement mode, the path cost value g(n) is described by g3(n) provided by the following formula (8):

[0070] g3(n) = l2·(ω l2 + ω b2 + ω t2 ·β) (8)

[0071] In the formula, ω l2 、ω b2 and ω t1 are respectively the path length penalty coefficient, reverse penalty coefficient, and wheel rotation angle penalty coefficient in the front and rear wheel same-direction movement mode. The numerical setting principle of the path cost value g(n) of the front and rear wheels in the same direction is relatively similar to that of the front and rear wheel different-direction movement mode, and can be slightly increased by 0.1 - 0.2 on the basis of the specific numerical values of ω l1 、ω b1 and ω t1 .

[0072] In one embodiment, in the case where the vehicle is in the front and rear wheel same-direction movement mode, the mode switching cost value p(n) is described as p3. To prevent unnecessary mode switching, and since the front and rear wheel different-direction movement mode includes in-situ rotation and requires stopping, it is relatively large. Therefore, the specific numerical value of p3 can be set in the range of 2 to 3.

[0073] Among them, the node expansion method in step S15 also adopts the node expansion method provided in the above embodiments.

[0074] In one embodiment, in step S16, according to whether the feasible child node is in the open list or the closed list, the following node information update is performed:

[0075] S161, if the child node n i+1 is not in the open list and the closed list, then set the parent node of this child node n i+1 to n i , and put the child node into the open list.

[0076] S162, if the child node n i+1 is in the open list, then this child node has been traversed. Obtain the old total cost value of the child node n i+1 in the open list, and compare it with the new total cost value calculated in step S15 with n i as the parent node. If the old total cost value is less than or equal to the new total cost value, then retain the information of this child node in the open list. If the old total cost value is greater than the new total cost value, it means that a better path can be obtained with n i as the parent node. At this time, replace the information of the child node n i+1 in the open list.

[0077] S163, if the child node is in the closed list, then skip this child node.

[0078] In one embodiment, in step S2, the path is segmented with the mode switching node on the global path as the end point. The path segments can be divided into two categories: the front and rear wheel reverse motion path segments and the front and rear wheel forward motion path segments. For the path segments generated by the front and rear wheel reverse motion mode, the path optimization is carried out using the gradient-based numerical optimization method; while the path segments generated by the front and rear wheel forward motion mode are straight lines and do not require path optimization. The path optimization includes the following steps:

[0079] S21, fully consider aspects such as path smoothness, curvature continuity, and collision safety, and design a comprehensive objective function that is convenient for solving the gradient;

[0080] S22, define the constraint conditions, including vehicle kinematic constraints, two-point boundary value constraints, and path constraints;

[0081] S23, when one end point of the path segment is a in-place steering node, the boundary value constraint of this end point can be simplified, and the body angle constraint of this end point can be removed;

[0082] S24, obtain the optimal weight coefficients of each item of the objective function through experimental analysis;

[0083] S25, obtain the final path through iterative optimization of the gradient information.

[0084] In this embodiment, the possible path quality of the front and rear wheel reverse motion mode path segments obtained through graph search may be relatively low. Taking this as the preliminary path, and then a path with higher quality can be obtained through path optimization.

[0085] Furthermore, in step S2, when performing speed planning on the path segments, the speeds at both ends of each path segment are constrained to zero, and the trapezoidal speed planning method is used to perform speed planning on the path to obtain the parking trajectory.

[0086] Such as Figure 6As shown in the figure, the present invention also provides a parking trajectory planning device for a four-wheel independent steering vehicle, which includes an environmental perception module, a parking space management module, a parking trajectory planning module, and a vehicle control module, where:

[0087] The environmental perception module has a perception function and is used to generate parking space information, a grid map, and vehicle initial position information. The environmental perception module is equipped with in-vehicle sensors such as cameras, ultrasonic radars, and lidar, which can identify parking spaces, perceive the vehicle pose and the surrounding environment, obtain parking space information, environmental information, and vehicle initial position information, and generate a grid map based on the environmental information. Finally, the environmental perception module transmits the parking space information, the grid map, and the vehicle initial position information to the parking space management module, and transmits the vehicle grid map and the initial position information to the parking trajectory planning module.

[0088] The parking space management module receives the parking space information, the grid map, and the vehicle initial position information transmitted by the environmental perception module, selects an optimal parking space that the vehicle can park in, and then generates target parking position information based on the selected parking space.

[0089] The parking trajectory planning module is used to receive the vehicle initial position information and the grid map transmitted by the environmental perception module and the target parking position information transmitted by the parking space management module, and uses the optimized global trajectory generated by the parking trajectory planning method of the four-wheel independent steering vehicle described in any one of claims 1-8 as the parking trajectory.

[0090] The vehicle control module is used to complete the parking task according to the trajectory information after receiving the parking trajectory information transmitted by the parking trajectory planning module. The vehicle control module has a control function and is used to control the vehicle to complete parking. After receiving the parking trajectory information transmitted by the parking trajectory planning module, the vehicle control module controls the steering wheel, throttle, brakes and other actuators of the vehicle according to the trajectory information to complete the parking task.

[0091] In summary, the complete working process of the entire parking trajectory planning system for a four-wheel independent steering vehicle is as follows:

[0092] 1) The parking trajectory planning system for a four-wheel independent steering vehicle is turned on;

[0093] 2) The environmental perception module perceives the surrounding environment, generates parking space information, a grid map, and vehicle initial position information, and transmits the information to the parking space management module and the parking trajectory planning module respectively;

[0094] 3) The parking space management module decides the optimal target parking position of the current vehicle according to the parking space information, the grid map, and the vehicle initial position information, and transmits the target parking position to the parking trajectory planning module;

[0095] 4) The parking trajectory planning module generates a parking trajectory using the above-mentioned four-wheel independent steering vehicle parking trajectory planning method based on the grid map, vehicle initial position, and target parking position information, and conveys the parking trajectory to the vehicle control module;

[0096] 5) The vehicle control module controls the vehicle to complete the parking task according to the parking trajectory information.

[0097] Regarding the technical solution in 2, are there any other alternative solutions that can also achieve the invention purpose?

[0098] Global path planning: First, identify special sections in the map, then use the standard hybrid A* algorithm to search for a path. When a special section is searched, switch to the in-place rotation or the same-direction movement mode of the front and rear wheels to pass through the specific section, and then continue to use the hybrid A* to search for the path, repeating until the path planning is completed.

[0099] Finally, it should be pointed out that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than limiting them. Those of ordinary skill in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some of the technical features can be equivalently replaced; these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A parking trajectory planning method for a four-wheel independent steering vehicle, characterized in that Including: S1. Obtain the initial position of the vehicle, the target parking position, and the grid map of the parking scenario. Using the target parking position as the search starting point and the initial position of the vehicle as the end point, perform global path planning with an improved hybrid A* algorithm to obtain the original global path. Among them, the improved hybrid A* algorithm includes a sub-node expansion step, and this sub-node expansion step adopts a sub-node expansion method corresponding to the vehicle motion mode preset in advance; S2. Using the mode switching nodes on the original global path as endpoints, segment the original global path, perform path optimization and speed planning on each segment of the path to obtain the optimized global trajectory; In step S1, the sub-node expansion method corresponding to the front and rear wheel reverse motion mode in the vehicle motion mode generates (k1 + 2) sub-nodes. The pose description of the left-rotating sub-node is expressed by the following formula (1), the pose description of the right-rotating sub-node is expressed by the following formula (2), and the pose description of the straight-line sub-node is expressed by the following formula (3): Wherein, (x1, y1) is the coordinate of one of the (k1 + 2) child nodes in the geodetic coordinate system, (x, y) is the coordinate of the parent node in the geodetic coordinate system, R is the turning radius, and i is all integers from 1 to k1 / 4, L is the wheelbase, θ is the maximum wheel steering angle, α is the vehicle heading angle at the parent node, Δα is the vehicle yaw angle, l1 is the step size, and α1 is the vehicle heading angle at one of the (k1 + 2) child nodes; The sub-node expansion method corresponding to the in-situ rotation motion mode in the vehicle motion mode generates k2 sub-nodes described by the following formula (5): In the formula, (x2, y2) is the coordinate of one of the k2 sub-nodes in the earth coordinate system, (x, y) is the coordinate of the parent node in the earth coordinate system, α is the vehicle heading angle at the parent node, i is all integers from 1 to k2, and α2 is the heading angle of one of the k2 sub-nodes; The sub-node expansion method corresponding to the front and rear wheel same-direction motion mode in the vehicle motion mode generates k3 sub-nodes described by the following formula (7): where \((x3, y3)\) is the coordinate of one of the \(k3\) child nodes in the geodetic coordinate system, \((x, y)\) is the coordinate of the parent node in the geodetic coordinate system, \(\alpha\) is the vehicle heading angle at the parent node, \(l2\) is the step size, \(\alpha3\) is the heading angle of one of the \(k3\) child nodes, and the front wheel steering angle \(j\) is all integers from 1 to \(k3 / 4\).

2. The parking trajectory planning method for a four-wheel independently-steering vehicle according to claim 1, characterized in that During the process of the "improved hybrid A* algorithm" in step S1, record the expansion method information of each generated node. When calculating the cost value of the sub-node, when the expansion method information of the parent node is the same as that of the sub-node, the total cost value of the sub-node is expressed as f(n) = g(n) + h(n); when the expansion method information of the parent node is different from that of the sub-node, the total cost value of the sub-node is expressed as f(n) = g(n) + h(n) + p(n), where f(n), g(n), h(n), and p(n) respectively represent the total cost value, path cost value, heuristic cost value, and mode switching cost value of node n.

3. The parking trajectory planning method for a four-wheel independent steering vehicle according to claim 2, wherein When calculating the heuristic cost value g(n) of the sub-node in the "improved hybrid A* algorithm" in step S1, when the expansion method information of the sub-node is the front and rear wheel reverse, in-situ rotation, and front and rear wheel same-direction motion modes respectively, the heuristic cost values g(n) are respectively described by g1(n), g2(n), and g3(n) provided by the following formulas: g1(n) = l1·(ω l1 + ω b1 + ω t1 ·Δα) (4) g2(n) = ω r ·|α2 - α| (6) g3(n) = l2·(ω l2 + ω b2 + ω t2 ·β) (8) where ω l1 , ω b1 and ω t1 are the path length penalty coefficient, reverse penalty coefficient, and turning penalty coefficient corresponding to the front and rear wheel different-direction movement mode expansion method, respectively; ω r is the rotation angle penalty coefficient corresponding to the in-place rotation mode expansion method; ω l2 , ω b2 and ω t2 are the path length penalty coefficient, reverse penalty coefficient, and wheel angle penalty coefficient corresponding to the front and rear wheel same-direction movement mode expansion method, respectively.

4. The parking trajectory planning method for a four-wheel independent steering vehicle according to claim 3, characterized in that, When calculating the mode switching cost value p(n) of the sub-node in the "improved hybrid A* algorithm" in step S1, if it is the mode switching between the front and rear wheel reverse and in-situ rotation motion modes, p(n) is described as a constant p1; if it is the mode switching between the front and rear wheel reverse and front and rear wheel same-direction motion modes, p(n) is described as a constant p2; if it is the mode switching between the front and rear wheel same-direction and in-situ rotation motion modes, p(n) is described as a constant p3.

5. The parking trajectory planning method for a four-wheel independent steering vehicle according to claim 1, characterized in that, In step S2, the original global path is segmented into a front and rear wheel reverse motion path segment and a front and rear wheel same-direction motion path segment, and a gradient-based numerical optimization method is used to optimize the path of the front and rear wheel reverse motion path segment.

6. A parking trajectory planning device for a four-wheel independent steering vehicle, characterized in that, Including: An environment perception module, which is used to generate parking space information, grid map and vehicle initial position information; A parking space management module, which is used to receive the parking space information, grid map and vehicle initial position information transmitted by the environment perception module, select an optimal parking space that the vehicle can park in, and then generate target parking position information according to the selected parking space; A parking trajectory planning module, which is used to receive the vehicle initial position information and grid map transmitted by the environment perception module and the target parking position information transmitted by the parking space management module, and use the optimized global trajectory generated by the parking trajectory planning method of the four-wheel independent steering vehicle described in any one of claims 1-5 as the parking trajectory; A vehicle control module, which is used to complete the parking task according to the trajectory information after receiving the parking trajectory information transmitted by the parking trajectory planning module. The vehicle control module has a control function for controlling the vehicle to complete parking; after receiving the parking trajectory information transmitted by the parking trajectory planning module, the vehicle control module controls the steering wheel, throttle, brake and other actuators of the vehicle according to the trajectory information to complete the parking task.

Images