Parking trajectory optimization method and vehicle

Abstract

The invention discloses a parking trajectory optimization method and a vehicle, and the method comprises the steps: deleting a short-section original trajectory with a higher curvature at a starting end, replacing the short-section original trajectory with a derivative trajectory section with a smaller curvature and a longer length, changing the geometric characteristics of the trajectory starting section, enabling the vehicle to start to move from a state with a smaller steering wheel angle, and improving the parking trajectory optimization efficiency. Smooth starting of the power steering wheel is achieved, and parking experience and efficiency are improved. By performing translation and rotation alignment based on the original starting point state on the generated initial sub-trajectory, it is ensured that the optimized trajectory starts from the actual starting position and posture of the vehicle, the accuracy of the parking starting point is ensured, and execution errors caused by trajectory deviation are avoided. Through fusion of the processed derivative sub-track and the original sub-track, the generated first fusion sub-track fits the optimized smooth characteristic at the starting part and returns to the accurate end point of the original track at the end part, and the positioning precision of the parking end point is kept while the whole process is ensured to be smooth and continuous.

Description

Technical Field

[0001] This invention relates to the field of automotive technology, and more particularly to a parking trajectory optimization method and a vehicle. Background Technology

[0002] During automatic parking, vehicles typically need to follow a planned trajectory. However, common parking trajectories often require a large steering wheel angle at the starting point, forcing the vehicle to first turn the steering wheel to a specific angle while stationary before it can begin to move. This process is not in line with human driving habits and is time-consuming. Improving the smoothness and consistency of the vehicle's parking initiation is a pressing issue that needs to be addressed. Summary of the Invention

[0003] This application provides a parking trajectory optimization method and vehicle, solving the technical problem in the prior art where the curvature of the parking trajectory at the starting point is not zero, causing the vehicle to need to turn the steering wheel while stationary to start or shift gears, resulting in a poor parking experience and low efficiency. It achieves the generation of a new trajectory with a smooth transition of curvature to zero at the starting point and shift point without changing the overall shape and end position of the parking trajectory. This allows the vehicle to move and turn (dynamically turn the steering wheel) and achieve dynamic and smooth gear shifting, effectively improving the naturalness, comfort and overall efficiency of parking operation.

[0004] Firstly, this application provides a parking trajectory optimization method, the method comprising: Obtain the original parking trajectory of the target vehicle at its current position. The original parking trajectory includes at least one original sub-trajectory, which refers to the trajectory segment between the positions of the target vehicle when the target vehicle changes gears at any two adjacent times. A first optimization step is performed on the starting end of the first segment of the original sub-trajectory of the original parking trajectory, where the current position is located, to obtain a first derived trajectory segment. The first segment of the original sub-trajectory is the target original sub-trajectory in the first optimization step, and the starting end of the first segment of the original sub-trajectory is the first end in the first optimization step. The first derived trajectory segment is combined with the trajectory segment retained in the first segment of the original sub-trajectory to obtain a first initial sub-trajectory. Based on the trajectory parameters of the trajectory feature point where the starting end of the first segment of the original sub-trajectory is located before the trajectory segment is deleted, the first initial sub-trajectory is translated and rotated so that the starting point of the first initial sub-trajectory coincides with the trajectory feature point where the starting end of the first segment of the original sub-trajectory is located before the trajectory segment is deleted, to obtain a first derived sub-trajectory. The first segment of the original sub-trajectory and the first derived sub-trajectory are fused to obtain a first fused sub-trajectory. If the original parking trajectory only includes the first segment of the original sub-trajectory, the first fused sub-trajectory is determined as the optimized target parking trajectory of the target vehicle. The first optimization step includes: deleting the original trajectory segment with a first preset length where the first end is located in the target original sub-trajectory, and then adding a derived trajectory segment with a second preset length to the first end of the target original sub-trajectory; the first preset length is less than the second preset length; and the curvature of the derived trajectory segment is less than the curvature of the original trajectory segment.

[0005] Secondly, this application provides a vehicle, including: processor; Memory used to store the processor's executable instructions; The processor is configured to execute a parking trajectory optimization method as provided in the first aspect.

[0006] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: The parking trajectory optimization method provided in this application achieves significant technical effects by acquiring an original parking trajectory containing at least one original sub-trajectory and performing a specific first optimization step on the starting end of the first original sub-trajectory. First, by deleting short segments of the original trajectory with high curvature at the starting end and replacing them with derived trajectory segments with lower curvature and longer lengths, the geometric characteristics of the trajectory's starting segment are fundamentally changed. This allows the vehicle to begin moving from a state with a smaller steering wheel angle, effectively solving the technical bottleneck of traditional parking where "turning the steering wheel while stationary" is required to start, achieving a smooth start with "moving the steering wheel," and improving parking experience and efficiency. Second, by aligning the generated initial sub-trajectory with translation and rotation based on the original starting point state, it ensures that the optimized trajectory starts from the vehicle's actual starting position and posture, guaranteeing the accuracy of the parking starting point and avoiding execution errors caused by trajectory deviation. Furthermore, by fusing the processed derived sub-trajectory with the original sub-trajectory, the generated first fused sub-trajectory adheres to the optimized smoothness characteristics at the beginning and reverts to the accurate endpoint of the original trajectory at the end, thus ensuring smooth continuity throughout while maintaining the positioning accuracy of the parking endpoint. In addition, this method adaptively outputs results by determining the number of trajectory segments. For simple parking scenarios containing only a single trajectory segment, this method can also effectively optimize the starting process, demonstrating the versatility and robustness of the solution. Therefore, this embodiment accurately optimizes the dynamic characteristics of the trajectory starting point without significantly altering the global shape of the original trajectory, balancing starting smoothness, operational naturalness, and parking endpoint accuracy. Attached Figure Description

[0007] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 A flowchart illustrating a parking trajectory optimization method provided in this application embodiment; Figure 2 A schematic diagram of the original parking trajectory provided in the embodiments of this application; Figure 3 A schematic diagram illustrating the principle of calculating the heading angle between two adjacent feature points, provided for an embodiment of this application; Figure 4 A schematic diagram illustrating the distribution of the first original sub-trajectory and the first derived sub-trajectory provided in an embodiment of this application; Figure 5 A schematic diagram showing the distribution of the first original sub-trajectory, the first derived sub-trajectory, and the first fused sub-trajectory provided in the embodiments of this application; Figure 6 This is a schematic diagram showing the end and start points of each sub-trajectory provided in the embodiments of this application after feature points have been added; Figure 7 A schematic diagram illustrating the distribution of the initial sub-trajectory at the front, the initial sub-trajectory at the rear, the front point, the rear endpoint, and the middle point, provided for an embodiment of this application. Figure 8 A schematic diagram showing the distribution of the initial sub-trajectory before the application, the initial sub-trajectory after the application, the intermediate point, the fused sub-trajectory before the application, and the fused sub-trajectory after the application; Figure 9 This is a structural schematic diagram of a vehicle provided in an embodiment of this application. Detailed Implementation

[0009] This application provides a parking trajectory optimization method and vehicle, which solves the technical problem in the prior art where the curvature of the parking trajectory is not zero at the starting point, causing the vehicle to need to turn the steering wheel while stationary to start or shift gears, resulting in a poor parking experience and low efficiency.

[0010] The technical solution of this application embodiment is to solve the above-mentioned technical problems, and the general idea is as follows: The parking trajectory optimization method provided in this application achieves significant technical effects by acquiring an original parking trajectory containing at least one original sub-trajectory and performing a specific first optimization step on the starting end of the first original sub-trajectory. First, by deleting short segments of the original trajectory with high curvature at the starting end and replacing them with derived trajectory segments with lower curvature and longer lengths, the geometric characteristics of the trajectory's starting segment are fundamentally changed. This allows the vehicle to begin moving from a state with a smaller steering wheel angle, effectively solving the technical bottleneck of traditional parking where "turning the steering wheel while stationary" is required to start, achieving a smooth start with "moving the steering wheel," and improving parking experience and efficiency. Second, by aligning the generated initial sub-trajectory with translation and rotation based on the original starting point state, it ensures that the optimized trajectory starts from the vehicle's actual starting position and posture, guaranteeing the accuracy of the parking starting point and avoiding execution errors caused by trajectory deviation. Furthermore, by fusing the processed derived sub-trajectory with the original sub-trajectory, the generated first fused sub-trajectory adheres to the optimized smoothness characteristics at the beginning and reverts to the accurate endpoint of the original trajectory at the end, thus ensuring smooth continuity throughout while maintaining the positioning accuracy of the parking endpoint. In addition, this method adaptively outputs results by determining the number of trajectory segments. For simple parking scenarios containing only a single trajectory segment, this method can also effectively optimize the starting process, demonstrating the versatility and robustness of the solution. Therefore, this embodiment accurately optimizes the dynamic characteristics of the trajectory starting point without significantly altering the global shape of the original trajectory, balancing starting smoothness, operational naturalness, and parking endpoint accuracy.

[0011] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0012] First, it should be noted that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, 1 and / or B can represent: 1 existing alone, 1 and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0013] This application provides a parking trajectory optimization method, which includes steps S11-S13, for details of which can be found in the following embodiments. Figure 1 As shown.

[0014] Step S11: Obtain the original parking trajectory of the target vehicle at its current position. The original parking trajectory includes at least one original sub-trajectory. The original sub-trajectory refers to the trajectory segment between the positions of the target vehicle when the target vehicle changes gears at any two adjacent times. Step S12: Perform a first optimization step on the starting end of the first segment of the original sub-trajectory of the original parking trajectory, where the current position is located, to obtain a first derived trajectory segment. The first segment of the original sub-trajectory is the target original sub-trajectory in the first optimization step, and the starting end of the first segment of the original sub-trajectory is the first end in the first optimization step. Combine the first derived trajectory segment with the trajectory segment retained in the first segment of the original sub-trajectory to obtain a first initial sub-trajectory. Based on the trajectory parameters of the trajectory feature point where the starting end of the first segment of the original sub-trajectory is located before the trajectory segment is deleted, translate and rotate the first initial sub-trajectory so that the starting point of the first initial sub-trajectory coincides with the trajectory feature point where the starting end of the first segment of the original sub-trajectory is located before the trajectory segment is deleted, to obtain a first derived sub-trajectory. Fuse the first segment of the original sub-trajectory and the first derived sub-trajectory to obtain a first fused sub-trajectory. Step S13: If the original parking trajectory only includes the first segment of the original sub-trajectory, the first fused sub-trajectory is determined as the optimized target parking trajectory of the target vehicle. The first optimization step includes: deleting the original trajectory segment with a first preset length where the first end is located in the target original sub-trajectory, and then adding a derived trajectory segment with a second preset length to the first end of the target original sub-trajectory; the first preset length is less than the second preset length; and the curvature of the derived trajectory segment is less than the curvature of the original trajectory segment.

[0015] The parking trajectory optimization method provided in this application can be executed by a vehicle, specifically by the vehicle's processor or by a cloud server connected to the vehicle. This application does not impose any restrictions on this method.

[0016] Regarding step S11, the original parking trajectory of the target vehicle at its current position is obtained. The original parking trajectory includes at least one original sub-trajectory, which refers to the trajectory segment between the positions of the target vehicle when the target vehicle changes gears at any two adjacent times.

[0017] The original parking trajectory can be generated in real time by the vehicle's path planning module using a specific path planning algorithm (such as the Hybrid A* search algorithm) based on environmental perception information (such as parking space and obstacle positions), or it can be retrieved from pre-stored standard parking trajectories. An original sub-trajectory is defined as the continuous path segment traversed by the vehicle between two adjacent changes in driving direction (i.e., gear shifting). For example, in a complete "forward-reverse-forward" parking process, each continuous forward or reverse path constitutes an original sub-trajectory.

[0018] Combination Figure 2The target vehicle is located at the starting point Q, where Q is the current position. The complete path from the starting point Q to the ending point Z is the original parking trajectory. In the original parking trajectory, the trajectory first moves forward (towards...) Figure 2 The vehicle travels from the top to point J1 (the first shift point), and this path QJ1 is the first original sub-track. Then it travels downwards and to the left to point J2 (the second shift point), and the path J1J2 is the second original sub-track. Next, it travels upwards and to the left to point J3 (the third shift point), and the path J3J4 is the third original sub-track. Finally, it travels downwards and to the left to the endpoint Z, and the path J4Z is the fourth original sub-track. Therefore, this original parking trajectory consists of four sequentially connected original sub-tracks, and the endpoints of each sub-track (J1, J3, J4, Z) are the positions where the vehicle shifts gears.

[0019] It is important to note that in the context of automatic parking, the vehicle is in its physical starting position (i.e., Figure 2 The initial state of point Q in the driving system is varied: it may be in drive (D) gear ready to move forward, or in reverse (R) gear ready to move backward, or even switch from park (P) gear to drive. Therefore, point Q itself does not necessarily involve a "gear shift" (i.e., a change in the direction of travel from forward to reverse).

[0020] To ensure the universality and consistency of the definition of the "original sub-trajectory," this application explicitly defines the logical starting point of the parking trajectory as the position where the first change in driving direction (gear shift) occurs. That is, regardless of the initial gear of the vehicle at the physical starting point Q, the algorithm begins identifying and processing the sub-trajectory from the starting point (i.e., the current position of the target vehicle). If the vehicle starts from point Q without shifting gears (e.g., moving directly forward), then the starting point of the first segment of the sub-trajectory is Q, but its status as the "first segment of the sub-trajectory" is defined by the first gear shift point (J1) thereafter. If the vehicle shifts gears at point Q (e.g., from R to D), then point Q itself is a gear shift point.

[0021] Regarding step S12, a first optimization step is performed on the starting end of the first segment of the original sub-trajectory of the original parking trajectory, where the current position is located, to obtain a first derived trajectory segment. The first segment of the original sub-trajectory is the target original sub-trajectory in the first optimization step, and the starting end of the first segment of the original sub-trajectory is the first end in the first optimization step. The first derived trajectory segment is combined with the trajectory segment retained in the first segment of the original sub-trajectory to obtain a first initial sub-trajectory. Based on the trajectory parameters of the trajectory feature point where the starting end of the first segment of the original sub-trajectory is located before the trajectory segment is deleted, the first initial sub-trajectory is translated and rotated so that the starting point of the first initial sub-trajectory coincides with the trajectory feature point where the starting end of the first segment of the original sub-trajectory is located before the trajectory segment is deleted, to obtain a first derived sub-trajectory. The first segment of the original sub-trajectory and the first derived sub-trajectory are fused to obtain a first fused sub-trajectory.

[0022] The first optimization step in this embodiment is described as follows: The first optimization step includes: Delete the original trajectory segment with a first preset length located at the first end of the target original sub-trajectory, and then add a derived trajectory segment with a second preset length at the first end of the target original sub-trajectory; the first preset length is less than the second preset length; the curvature of the derived trajectory segment is less than the curvature of the original trajectory segment.

[0023] In the first optimization step, the derived trajectory segment is designed to have a smaller curvature than the deleted original trajectory segment. Specifically, the original trajectory segment often has a high curvature (i.e., a large steering angle) near the starting point or shift point, causing the vehicle to have to turn the steering wheel almost in place to start moving along the trajectory. The newly added derived trajectory segment extends smoothly forward or backward, with its path curvature starting from near zero (in the case of starting point extension) or gradually transitioning to zero (in the case of shift point), resulting in an overall curvature significantly lower than the original segment. The advantage of this design is that it allows the vehicle to start moving from a state where the steering wheel angle is close to the middle (small curvature), thereby achieving a smooth start or dynamic shift while "steering in motion," effectively eliminating the lag when turning the steering wheel in place, and improving the naturalness and overall efficiency of parking operations.

[0024] Furthermore, the first optimization step may specifically include steps S121-S124.

[0025] Step S121: Divide the original target sub-trajectory into N-1 trajectory units with a preset sub-length. The endpoints at both ends of each trajectory unit are denoted as trajectory feature points. The original target sub-trajectory includes N trajectory feature points; N is a positive integer.

[0026] Step S121 defines the basic operations for trajectory discretization. The "preset sub-length" can be set according to actual needs and computational accuracy; for example, a common and effective value is 0.05 meters (i.e., 5 centimeters). The core of this step is to discretize a continuous original sub-trajectory into a series of sequentially connected "trajectory units" by sampling at fixed intervals along this fixed length. Each unit has an equal length, and the connection points at both ends are defined as "trajectory feature points." If a sub-trajectory is divided into N-1 units, it will contain N feature points, providing a precise coordinate framework and data index for subsequent steps involving positioning, deletion, and extension based on these feature points.

[0027] Step S122: Starting from the trajectory feature point of the target original sub-trajectory at the first end, delete the original trajectory segment with a first preset length. The original trajectory segment refers to the trajectory segment composed of all trajectory units between the first trajectory feature point and the (M+1)th trajectory feature point, where M is a positive integer and 2 < M < N.

[0028] In step S122, the "first end" refers to the end to be processed. It can be either the starting end (the first endpoint passed in the parking sequence) or the ending end (the last endpoint passed in the parking sequence), depending on which segment of the original parking trajectory the target original sub-trajectory being processed is. The starting and ending ends are defined based on the order in which the vehicles travel along the parking trajectory. The core operation of this step is to delete a continuous trajectory segment starting from the "first end" and either backward (if the first end is the starting end) or forward (if the first end is the ending end). Specifically, it starts from the first trajectory feature point (located at the first end), counts backward to the (M+1)th trajectory feature point, and deletes the entire original trajectory segment consisting of these M trajectory units (i.e., the M+1 feature points). For example, if M=4, it means that starting from the first feature point of the first end, the original trajectory segment corresponding to a total of 4 trajectory units (containing 5 feature points) between it and the subsequent 5th feature point is deleted, thereby creating an operation space for the subsequent smooth extension of the "first end" of the target original sub-trajectory.

[0029] Step S123: Based on the trajectory parameters corresponding to the (M+1)th trajectory feature point, determine K derived feature points on the side away from the (M+1)th trajectory feature point. Among these, there is a preset sub-length distance between each pair of adjacent feature points from the (M+1)th trajectory feature point to the Kth derived feature point; the sum of the straight-line distances between each pair of adjacent feature points from the (M+1)th trajectory feature point to the Kth derived feature point is the second preset length; the difference in steering wheel angles between any two adjacent feature points from the first derived feature point to the (K-1)th derived feature point is the same; the steering wheel angles of each feature point from the (K-1)th derived feature point to the Kth derived feature point are 0; the vehicle heading angles of each feature point from the (K-2)th derived feature point to the Kth derived feature point are the same; K is a positive integer, K > 4, and M < K < N. Regarding step S123, based on the state of the deleted breakpoint (the (M+1)th trajectory feature point), K new derived feature points are sequentially generated on its outer side (in the direction away from the original trajectory) to construct a new trajectory to replace the deleted segment. The core technology lies in designing precise generation rules, ensuring that the steering wheel angle of the newly added points decreases uniformly from the original value until the last two points reach zero, while maintaining consistent vehicle heading angles at the last three points. This generates a derived trajectory segment with curvature that smoothly transitions from the original value to zero. This step creates a gently curvature, linearly decaying "introduction segment" or "transition segment" for the original trajectory, which originally had high curvature at the starting point or shift point. This allows the vehicle to follow the trajectory from a small steering wheel angle or a straightened state, providing the necessary smooth path foundation for dynamic starts or dynamic shifts that allow for "movement and steering simultaneously."

[0030] Specifically, the trajectory parameters include trajectory coordinates, vehicle heading angle, steering wheel angle, and driving direction. Step S123 may include steps S1231-S1233.

[0031] Step S1231: Determine the steering wheel angle of the (M+1)th trajectory feature point as the steering wheel angle of the first derived feature point; the difference between the steering wheel angles of any two adjacent feature points from the first derived feature point to the (K-1)th derived feature point is the same, the steering wheel angles of the (K-1)th derived feature point and the Kth derived feature point are 0, and the vehicle heading angles of all feature points from the (K-2)th derived feature point to the Kth derived feature point are the same.

[0032] In the original parking trajectory, each trajectory point (including trajectory feature points) contains a complete set of trajectory parameters used to accurately describe the vehicle's state at that point. Trajectory coordinates typically refer to the position of the vehicle's rear axle center point in a global coordinate system (such as the world coordinate system or parking lot coordinate system), denoted by (x, y). The vehicle heading angle is the angle between the vehicle's longitudinal axis and the positive X-axis of the global coordinate system, usually expressed in radians, representing the direction the vehicle is facing. The steering wheel angle is the turning angle of the vehicle's front wheels relative to the vehicle's centerline; its sign usually indicates turning left or right, and this angle directly determines the vehicle's instantaneous turning radius. The driving direction refers to the vehicle's gear direction at that point, usually simplified to "forward" (corresponding to drive gear) or "backward" (corresponding to reverse gear), used to determine the direction of trajectory extension.

[0033] Regarding step S1231, its core is determining the steering wheel angle sequence of the derived feature points. First, the steering wheel angle of the first derived feature point is inherited from the steering wheel angle of the (M+1)th trajectory feature point. This inheritance ensures the continuity and lack of abrupt changes in the motion state of the derived trajectory segment and the original trajectory segment at the connection point. The (M+1)th trajectory feature point is the remaining breakpoint after deleting the original trajectory segment and also the starting reference point for generating the derived trajectory; its steering wheel angle reflects the instantaneous steering state required by the vehicle at that position. If the steering wheel angle of the first derived point is not consistent with it, a step change in steering wheel angle will occur at the connection point, resulting in discontinuous trajectory curvature. The vehicle must then suddenly adjust its steering to follow the trajectory, thus disrupting smoothness. By inheriting the same steering wheel angle, the derived trajectory can start from the current steering state of the original trajectory and then achieve a smooth and continuous transition of curvature from the original value to zero through a uniform decrease in the angle of subsequent points. This is the basis for realizing "dynamic steering wheel control" or dynamic gear shifting.

[0034] Then, the steering wheel angle decreases point by point with a fixed difference (step size) until it decreases to 0 at the (K-1)th point, and the Kth point also remains at 0. The formula for calculating the steering wheel angle difference (step size) is: delta_steer = initial steering wheel angle / (K-2). For example: assuming the steering wheel angle at the (M+1)th point is 30 degrees and K=8, then delta_steer = 30° / (8-2) = 5°. Therefore, the steering wheel angle at the first derived feature point is 30°, the second is 25°, the third is 20°, and so on, with the sixth (K-1)th point being 5°, the seventh (K-1)th point being 0°, and the eighth (K)th point also being 0°.

[0035] Step S1232, i iterates sequentially from 1 to K-1, and performs the first step for i, which includes: Based on the steering wheel angle at the i-th derived feature point and the wheelbase of the target vehicle, the i-th driving radius of the target vehicle at the i-th derived feature point is determined; Based on the preset sub-length between the i-th derived feature point and the (i-1)-th derived feature point and the i-th driving radius, the i-th difference in vehicle heading angle between the i-th derived feature point and the (i-1)-th derived feature point is determined; Based on the vehicle heading angle of the (i-1)th derived feature point and the i-th difference, determine the vehicle heading angle of the i-th derived feature point; Based on the i-th driving radius, the trajectory coordinates of the (i-1)-th derived feature point, and the i-th difference, determine the trajectory coordinates of the i-th derived feature point.

[0036] Regarding step S1232, its "first step" is based on the vehicle kinematics model (bicycle model) to calculate the position and heading changes from the directional control quantity (steering wheel angle). The specific principle is as follows: Determine the driving radius: Using the formula R=wheel_base / tan(steer), calculate the instantaneous turning radius R at the current point based on the steering wheel angle steer and the vehicle wheelbase wheel_base. For example: if the wheelbase is 2.8 meters and the steering wheel angle is 30° (approximately 0.5236 radians), then R≈2.8 / tan(0.5236)≈4.85 meters.

[0037] Calculate the heading angle difference: When a vehicle travels along an arc of radius R for a small preset sub-length ds (e.g., 0.05 meters), the change in its heading angle is delta_yaw = ds / R. Figure 3 As shown, the line connecting feature point [i-1] and feature point i is ds, and the dot is the center of the circle corresponding to the current driving radius R. Continuing the previous example: ds=0.05, then delta_yaw≈0.05 / 4.85≈0.0103 radians.

[0038] Update the heading angle: The new heading angle is equal to the heading angle of the previous point plus (or minus, depending on the direction of travel) the calculated delta_yaw. Continuing the previous example: if the heading angle of the previous point is 1.0 radians, then the heading angle of the new point is approximately 1.0 + 0.0103 = 1.0103 radians.

[0039] Calculate trajectory coordinates: In the vehicle's local coordinate system, based on the geometric relationship of circular motion, calculate the coordinate offsets delta_x and delta_y from the old point to the new point (refer to...). Figure 3 As shown in the figure, the coordinates are then transformed to the global coordinate system and superimposed on the previous point coordinates to obtain the new point coordinates.

[0040] Step S1233: After i takes K-1 and executes the first step, the trajectory coordinates of the K-1 derived feature point are determined based on the trajectory coordinates of the K-1 derived feature point, the vehicle heading angle, and the preset sub-length between the K-1 derived feature point and the K-1 derived feature point.

[0041] Regarding step S1233, the principle is that after generating the (K-1)th point, since the steering wheel angle is already 0, the vehicle enters a straight-line driving state. At this time, the heading angle no longer changes, so the heading angle of the Kth point is exactly the same as that of the (K-1)th point. The calculation of its trajectory coordinates is simplified as follows: based on the position of the (K-1)th point, move forward or backward (determined by the driving direction) by a preset sub-length ds straight-line distance along the heading angle direction of that point. For example: assuming the coordinates of the (K-1)th point are (10.0, 5.0), the heading angle is 0.5 radians, and ds = 0.05 meters. Then the heading angle of the Kth point is still 0.5 radians, and its coordinates are calculated as: x_K = 10.0 + 0.05 * sin(0.5), y_K = 5.0 + 0.05 * cos(0.5).

[0042] Step S124: Based on the (M+1)th trajectory feature point to the Kth derived feature point, obtain the derived trajectory segment.

[0043] Step S124 connects and integrates the KM derived feature points (starting from the (M+1)th trajectory feature point and ending at the Kth derived feature point) generated in step S123 according to their generation order, thereby constructing a complete new trajectory to replace the deleted original trajectory segment, i.e., the derived trajectory segment. This step not only completes the morphological construction of the derived point sequence into a continuous trajectory, but more importantly, it ensures that the state (coordinates, heading angle, steering wheel angle) of the new trajectory at the starting end (i.e., at the (M+1)th point) is completely consistent with the break point of the original trajectory, achieving seamless connection. At the same time, by following the preset rules of decreasing angle and straight line generation, the derived trajectory segment naturally has the characteristic that the curvature continuously and smoothly decreases from the initial value until it returns to zero. Therefore, the final output of step S124 is a key trajectory segment with the expected smoothness that can be directly used for subsequent alignment and fusion operations, providing a direct path basis for eliminating stationary steering wheel operations.

[0044] Regarding the part in step S12 where operations are performed on the starting end of the first original sub-trajectory to obtain the first derived trajectory segment, its specific execution process relies entirely on the aforementioned first optimization step (i.e., steps S121 to S124). This process follows the rules defined in the first optimization step, including trajectory discretization, deletion of the original segment, generation of derived feature points, and construction of the derived trajectory segment. Since the specific content of the first optimization step has been explained in detail above, for the sake of brevity, its implementation details will not be repeated here.

[0045] The fundamental reason for combining the first derived trajectory segment with the retained original trajectory segment in step S12, and aligning its starting point with the original starting point through translation and rotation, lies in the fact that the initial state of the vehicle when it begins parking (i.e., its position, heading angle, etc., at the beginning of the original trajectory) is a physical constraint determined by global path planning and cannot be changed. Since the first optimization step generates the derived trajectory segment based on the new breakpoint (the (M+1)th trajectory feature point) after deleting the original segment, the starting point of the generated "first initial sub-trajectory" is spatially offset from the vehicle's actual starting position. To ensure the optimized trajectory can be executed by the vehicle from the correct initial state, a rigid body transformation involving translation and rotation must be applied to the entire "first initial sub-trajectory." The goal of this transformation is to ensure that the starting point of the transformed trajectory (i.e., the first derived sub-trajectory) completely coincides with the starting point of the original parking trajectory before the deletion operation in terms of position and heading angle, thereby providing the vehicle with a more smoothly curvatured drivable path starting from that point while maintaining a consistent initial state. For example, as... Figure 4 As shown, the derived trajectory segment is... Figure 4 For the points in part D, L1 is the first original sub-trajectory, and L2 is the first derived sub-trajectory.

[0046] Regarding step S12, which involves fusing the first original sub-trajectory and the first derived sub-trajectory to obtain the first fused sub-trajectory, linear averaging fusion can be used. The process is typically as follows: First, the two trajectories to be fused (here, the first original sub-trajectory and the first derived sub-trajectory) are resampled using linear interpolation to form a sequence with the same number of trajectory points, where these points correspond one-to-one according to their cumulative travel distance. Then, for each pair of corresponding trajectory points, the arithmetic mean of their various state parameters (such as coordinates x, y, and heading angle yaw) is directly taken as the parameter of the new fused trajectory at that point. The specific formula is: Fusion point parameter = (Original trajectory point parameter + Derived trajectory point parameter) / 2. Finally, all the averaged new points are connected sequentially to form the fused trajectory.

[0047] This method is simple in principle and requires little computation, but its main drawback is that it uniformly mixes the features of the two input trajectories along the entire trajectory. Since the two trajectories usually have significant morphological differences near the start and end points, this uniform mixing will result in the fused trajectory not completely matching either the original trajectory or the derived trajectory at the start and end points, potentially leading to undesirable position or heading deviations.

[0048] In addition, regarding the fusion of the first original sub-trajectory and the first derived sub-trajectory in step S12 to obtain the first fused sub-trajectory, the following steps S125-S129 can also be used.

[0049] Step S125: Divide the first segment of the original sub-trajectory into N-1 first trajectory units with a preset sub-length. The endpoints of each first trajectory unit are denoted as first trajectory feature points. Each first trajectory feature point includes a first cumulative length value, which refers to the sum of the lengths of all first trajectory units before the first trajectory feature point. The first segment of the original sub-trajectory includes N first trajectory feature points; N is a positive integer.

[0050] Regarding step S125, its core is to perform standardized discretization processing based on a fixed length on the first segment of the original sub-trajectory. This step first cuts the continuous original sub-trajectory into multiple equal-length "first trajectory units," each unit having a preset sub-length (e.g., 0.05 meters). The connection points between units are called "first trajectory feature points." Simultaneously, a key "first cumulative length value" is calculated for each feature point, representing the sum of the lengths of all units traversed from the trajectory's starting point to that feature point. Through this processing, the original continuous trajectory is transformed into a sequence of N discrete feature points. Each point not only contains state parameters such as position and heading but also includes precise mileage information. This lays the foundational data structure for subsequent fusion calculations with the derived trajectory based on the same number of points and precise positional correspondence.

[0051] Step S126: Divide the total length of the first derived sub-trajectory into N-1 second trajectory units. The endpoints of each second trajectory unit are denoted as second trajectory feature points. Each second trajectory feature point includes a second cumulative length value, which is the sum of the lengths of all second trajectory units before the second trajectory feature point. The first derived sub-trajectory includes N second trajectory feature points.

[0052] Regarding step S126, its core is to perform equal-point discretization processing on the first derived sub-trajectory based on the total length. This step does not directly use a preset sub-length for division; instead, it first divides the total length of the derived trajectory (derived total length) into N-1 equal parts, thus obtaining N "second trajectory feature points." This means that although the length of the derived trajectory differs from the original trajectory, through this operation, it is sampled as a sequence of feature points with the exact same number (N) as the original trajectory. Each second feature point also calculates a "second cumulative length value," representing the normalized mileage from the starting point of the derived trajectory to that point. The key purpose of this step is to achieve alignment of the two trajectories in the "point count" dimension, ensuring that in steps S127 and S128, interpolation and fusion calculations can be performed based on one-to-one corresponding points, thereby overcoming the problem that the two trajectories are difficult to directly fuse due to differences in length and shape.

[0053] Step S127: Using each first trajectory feature point as a reference, determine the first deviation degree of each second trajectory feature point relative to the first trajectory feature point, and perform interpolation processing on each second trajectory feature point based on the first deviation degree to obtain N first target interpolation points.

[0054] Regarding step S127, a common interpolation method is nearest neighbor interpolation. Its process is relatively simple and direct: when it's necessary to find a corresponding interpolation point on the second trajectory for a specific feature point (reference point) on the first trajectory, the algorithm iterates through all feature points on the second trajectory, directly calculating the Euclidean distance between each point and the current reference point. Then, it simply selects the feature point on the second trajectory with the smallest distance, directly copying its coordinates, heading angle, and other state parameters as the required interpolation result. This method completely ignores the cumulative distance relationship and path shape between trajectory points, relying solely on spatial proximity for a crude matching. This may lead to incorrect trajectory extension order of the interpolated points and the inability to generate new points located between known feature points. Therefore, the fusion result is usually rather rigid and inaccurate.

[0055] Furthermore, step S127 can also be implemented in the following way: Each first trajectory feature point in the first original sub-trajectory is sequentially taken as a first target feature point and the second step is executed to obtain N first target interpolation points corresponding to the first derived sub-trajectory. The second step includes: From the feature points in the first derived sub-trajectory whose second cumulative length value is greater than the first cumulative length of the first target feature point, select the second target feature point with the smallest second cumulative length value; From the feature points in the first derived sub-trajectory whose second cumulative length value is less than the first cumulative length of the first target feature point, select the third target feature point with the largest second cumulative length value; Based on the first cumulative length value of the first target feature point, the second cumulative length value of the second target feature point, and the second cumulative length value of the third target feature point, a first degree of deviation between the second target feature point and the third target feature point is determined. Based on the trajectory parameters of the second target feature point, the trajectory parameters of the third target feature point, and the first deviation degree, the first target interpolation point corresponding to the second target feature point and the third target feature point is determined.

[0056] This embodiment of the application focuses on the original trajectory, precisely calculating a corresponding position on the derived trajectory for each feature point (the first target feature point). Specifically, the algorithm traverses the N feature points of the original trajectory, performing the same query and calculation process (i.e., the "second step") for each point. This process locates the most suitable "target interval" on the derived trajectory based on the cumulative length value of the current original trajectory points, and generates a new point within that interval through interpolation. Ultimately, a corresponding interpolation point is generated for each point of the original trajectory, resulting in a set of N "first target interpolation points" that correspond one-to-one with the original trajectory points. This ensures that the contributions of the two trajectories can be precisely combined point-to-point during subsequent fusion.

[0057] The following explains each sub-step in "Step Two": Filtering the second target feature point: This step finds a position in the first derived sub-trajectory that "just exceeds" the current mileage of the first target feature point. Specifically, it filters all derived trajectory feature points whose cumulative length value is greater than the cumulative length value of the first target feature point, and selects the point with the smallest cumulative length value. This point can be regarded as a "forward sentinel" on the derived trajectory that first "arrives or crosses" the current mileage of the original trajectory.

[0058] Filtering the third target feature point: This step finds a position that "just hasn't exceeded" the current mileage of the first target feature point. It filters out all derived trajectory feature points whose cumulative length value is less than that of the first target feature point, and selects the point with the largest cumulative length value. This point can be considered a "rear sentinel" on the derived trajectory that is closest to but has not yet "reached" the current mileage of the original trajectory. The second and third target feature points together define a line segment interval on the derived trajectory that includes the target position.

[0059] Determine the first degree of deviation: This step calculates the positional proportion of the current first target feature point relative to the derived trajectory interval found above. The formula is: First degree of deviation = (Cumulative length value of the first target feature point - Cumulative length value of the third target feature point) / (Cumulative length value of the second target feature point - Cumulative length value of the third target feature point). The calculation result is a proportional coefficient between 0 and 1, which precisely represents the relative position of the target point on the line segment formed by the third and second target feature points (0 indicates coincidence with the third point, 1 indicates coincidence with the second point).

[0060] Determine the first target interpolation point: This step, based on the calculated first deviation degree, performs linear interpolation on various trajectory parameters (such as coordinates and heading angle) of the second and third target feature points. For example, the X-coordinate of the interpolation point = X-coordinate of the third target feature point + first deviation degree × (X-coordinate of the second target feature point - X-coordinate of the third target feature point). By performing this calculation for each parameter, a "first target interpolation point" that precisely corresponds to the current point of the original trajectory in both position and state is finally generated.

[0061] Step S128: Using each second trajectory feature point as a reference, determine the second deviation degree of each first trajectory feature point relative to the second trajectory feature point, and perform interpolation processing on each first trajectory feature point based on the second deviation degree to obtain N second target interpolation points; the N first target interpolation points and the N second target interpolation points correspond one-to-one along the extension direction of the first segment of the original sub-trajectory.

[0062] Regarding step S128, another conventional interpolation method is fixed-step resampling interpolation. This process does not rely on the mutual reference relationship between the two trajectories, but rather performs independent preprocessing on the second trajectory (here, the first trajectory): starting from the trajectory starting point according to a pre-set fixed physical length (e.g., 0.1 meters), a series of new, equally spaced sampling points are generated along the trajectory path. When it is necessary to find the corresponding point on the first trajectory for a certain feature point on the second trajectory, the algorithm will directly select the point at the same proportion position in the resampling sequence of the first trajectory according to the normalized proportion of that point on the second trajectory (e.g., mileage percentage), or perform simple linear interpolation on two adjacent resampling points. Although this method ensures the uniform distribution of its own trajectory points, it ignores the correspondence logic with the points of the other trajectory. The interpolated points of the two trajectories lack a correspondence based on a common reference, which may lead to systematic deviations or distortions during fusion. Steps S127 and S128 can respectively adopt the same or different interpolation methods of nearest neighbor interpolation and fixed-step resampling interpolation. This application embodiment does not limit or elaborate on this.

[0063] Furthermore, step S128 can also be implemented in the following way: Each second trajectory feature point in the first derived sub-trajectory is sequentially used as the fourth target feature point and the third step is executed to obtain N second target interpolation points corresponding to the first original sub-trajectory. The third step includes: From the feature points in the first original sub-trajectory segment whose first cumulative length value is greater than the second cumulative length of the fourth target feature point, select the fifth target feature point with the smallest first cumulative length value. From the feature points in the first original sub-trajectory whose first cumulative length value is less than the second cumulative length of the fourth target feature point, select the sixth target feature point with the largest first cumulative length value; Based on the second cumulative length value of the fourth target feature point, the first cumulative length value of the fifth target feature point, and the first cumulative length value of the sixth target feature point, a second degree of deviation between the fifth target feature point and the sixth target feature point is determined. Based on the trajectory parameters of the fifth target feature point, the trajectory parameters of the sixth target feature point, and the second deviation degree, the second target interpolation point corresponding to the fifth target feature point and the sixth target feature point is determined.

[0064] Steps S128 and S127 are symmetrical and complementary, with the derived trajectory as the main driver, precisely calculating a corresponding position for each feature point (the fourth target feature point) on the original trajectory. Specifically, the algorithm traverses the N feature points of the derived trajectory, performing the same query and calculation process (i.e., "step three") for each point. This process locates the most suitable "target interval" on the original trajectory based on the cumulative length value of the current derived trajectory points, and generates a new point within that interval through interpolation. Finally, a corresponding interpolation point is generated for each point of the derived trajectory, resulting in another set of N "second target interpolation points" that correspond one-to-one with the derived trajectory points in sequence. This step, together with step S127, ensures that a bidirectional and precise correspondence is established between the points on the two trajectories, laying the data foundation for subsequent point-by-point fusion.

[0065] The following explains each sub-step in "Step Three": Selecting the fifth target feature point: This step involves finding a position within the first original sub-track that "just exceeds" the mileage of the current fourth target feature point (derived trajectory point). Specifically, it filters all original trajectory feature points whose first cumulative length value is greater than the second cumulative length value of the fourth target feature point, and selects the point with the smallest first cumulative length value (the fifth target feature point). This point can be considered a "forward sentinel" on the original trajectory that first "arrives at or surpasses" the mileage of the current point on the derived trajectory.

[0066] Filtering the sixth target feature point: This step finds a position that "just hasn't exceeded" the current mileage of the fourth target feature point. It filters out all original trajectory feature points whose first cumulative length value is less than the second cumulative length value of the fourth target feature point, and selects the point with the largest first cumulative length value (the sixth target feature point). This point can be considered a "rear sentinel" on the original trajectory that is closest to but has not yet "reached" the current mileage of the derived trajectory. The fifth and sixth target feature points together define a line segment interval on the original trajectory that contains the target location.

[0067] Determine the second deviation degree: This step calculates the positional proportion of the current fourth target feature point relative to the original trajectory interval found above. The formula is: Second deviation degree = (Second cumulative length value of the fourth target feature point - First cumulative length value of the sixth target feature point) / (First cumulative length value of the fifth target feature point - First cumulative length value of the sixth target feature point). The calculation result is a proportional coefficient between 0 and 1, which precisely represents the relative position of the current point of the derived trajectory on the original trajectory line segment formed by the sixth and fifth target feature points.

[0068] Determine the second target interpolation point: This step involves linearly interpolating the trajectory parameters of the fifth and sixth target feature points based on the calculated second deviation degree. For example, the X-coordinate of the interpolation point = X-coordinate of the sixth target feature point + second deviation degree × (X-coordinate of the fifth target feature point - X-coordinate of the sixth target feature point). By performing this calculation for each parameter, a "second target interpolation point" that precisely corresponds to the current point of the derived trajectory in both position and state is ultimately generated.

[0069] Step S129: Based on N first target interpolation points and N second target interpolation points, obtain N fusion feature points, and determine the first fusion sub-trajectory based on the N fusion feature points.

[0070] Regarding step S129, a conventional fusion method can be used: fixed-weight endpoint constraint fusion. This method first ensures that the two trajectories (the first target interpolation point sequence and the second target interpolation point sequence) have the same number of points N and correspond one-to-one. Then, it assigns a fixed fusion weight (e.g., 0.5) to the first N-1 point pairs excluding the endpoint, and calculates the parameters of the fusion point through a weighted average: Fusion point parameter = weight * first target interpolation point parameter + (1 - weight) * second target interpolation point parameter. For the last point (i.e., the endpoint), it is mandatory that all parameters of the fusion point must be exactly equal to the endpoint parameters of the first original sub-trajectory segment, thus ensuring the accuracy of the final parking position. Although this method satisfies the position requirements through hard endpoint constraints, its fixed-weight design cannot implement differentiated fusion strategies in different parts of the trajectory, resulting in potentially insufficient smoothness in the initial segment and weak overall trajectory smoothness.

[0071] In addition, step S129 can also be implemented in the following manner, including steps S1291-S124.

[0072] Step S1291: Take the first a first target interpolation points out of N first target interpolation points as the first a fusion feature points of the first fusion sub-trajectory; take the last b second target interpolation points out of N second target interpolation points as the last b fusion feature points of the first fusion sub-trajectory; a and b are positive integers, a≤8, b≤8.

[0073] Step S1291 defines the processing strategy for the start and end segments of the fused trajectory. It directly uses the first *a* points of the derived trajectory (the first target interpolation point sequence after interpolation alignment) as the starting segment of the fused trajectory, and the last *b* points of the original trajectory (the second target interpolation point sequence after interpolation alignment) as the ending segment of the fused trajectory. Parameters *a* and *b* are positive integers, typically small (e.g., less than or equal to 8). The core purpose of this design is to: at the beginning of the trajectory, force the fused result to completely fit the derived trajectory, thus ensuring the vehicle can start from the smoothly curvatured extension segment, achieving "dynamic steering"; at the end of the trajectory, force the fused result to completely revert to the original trajectory, thus ensuring the parking endpoint's position and heading are consistent with the original plan, without any pose deviation. This sets clear boundary constraints for the entire trajectory.

[0074] "Moving the steering wheel" specifically refers to the process of smoothly and continuously turning the steering wheel to follow the target trajectory while the vehicle begins to move along the parking trajectory or after changing direction at a gear shift point, while the vehicle body undergoes substantial movement (i.e., not stationary). Its technical implications directly contrast with the traditional "turning the steering wheel while stationary" operation commonly used in parking. Traditional mode (steering while stationary): The vehicle is completely stationary at the starting point or shift point, and the steering wheel needs to be turned quickly to a large predetermined angle before the vehicle begins to move. This results in jerky starts, longer driving times, and does not conform to human driving habits.

[0075] This solution (dynamic steering wheel): By smoothly extending and optimizing the starting point and shift point of the original trajectory, a new trajectory with near-zero curvature at the beginning (steering wheel near the center position) is generated. The vehicle starts moving from a small steering wheel angle, and during the movement, the steering wheel angle increases synchronously and continuously with the smooth increase of the trajectory curvature until the required steering angle is reached.

[0076] The core technology behind this effect lies in trajectory preprocessing: by deleting segments with high curvature at the beginning of the original trajectory and replacing them with a derived trajectory whose curvature smoothly increases from zero, the vehicle can start from a near-straight position, thus achieving a natural driving experience of "moving and turning simultaneously." This not only improves the comfort and efficiency of the parking process but also reduces tire wear and the load on the steering system.

[0077] Step S1292: For the (a+1)th to the Nbth first target interpolation points among the N first target interpolation points, and the (a+1)th to the Nbth second target interpolation points among the N second target interpolation points, determine the first fusion coefficient between each pair of one-to-one corresponding first target interpolation points and second target interpolation points based on the first formula; the first formula is: coef=0.5*{1-cos[(ia)*PI / (Nab)]}; Wherein, coef is the first fusion coefficient of the i-th corresponding first target interpolation point and second target interpolation point among the N first target interpolation points and N second target interpolation points; cos[(ia)*PI / (Nab)] represents the cosine function of (ia)*PI / (Nab); i represents the index of the i-th interpolation point in the extension direction along the first original sub-trajectory from the starting point to the ending point among the N first target interpolation points and N second target interpolation points, i is a positive integer, and a<i<N-b+1; PI represents π.

[0078] Step S1292 is responsible for calculating a dynamically changing fusion weight (first fusion coefficient coef) for the intermediate transition between the starting and ending segments. It uses a formula based on a cosine function: coef = 0.5 * {1 - cos[(ia) * PI / (Nab)]}. In this formula, i is the index of the current point in the overall sequence. The ingenuity of this formula lies in the fact that as i increases from a+1 to Nb, the value of (ia) * PI / (Nab) smoothly increases from 0 to PI, causing coef to correspondingly and smoothly increase from 0 to 1. This change curve has a zero rate of change at the starting point (coef = 0) and the ending point (coef = 1), achieving a smooth "gradual entry and exit" transition effect. This avoids the abrupt changes in fusion parameters that might be caused by linear weight changes, thus ensuring the smoothness and continuity of the intermediate segment's fusion trajectory.

[0079] Step S1291: Based on the (a+1)th to Nbth first target interpolation points among the N first target interpolation points, the (a+1)th to Nbth second target interpolation points among the N second target interpolation points, and the first fusion coefficients corresponding to each pair of first target interpolation points and second target interpolation points, determine the (a+1)th to Nbth fusion feature points in the first fusion sub-trajectory. For details, refer to the following second formula: point_merge=coef*P1+(1-coef)*P2; Where point_merge represents any fusion trajectory point from the (a+1)th fusion feature point to the Nbth fusion feature point, coef represents the first fusion coefficient, P1 represents the second target interpolation point, and P2 represents the first target interpolation point.

[0080] Step S1293, based on the dynamic fusion coefficient *coef* calculated in step S1292, performs weighted fusion on each pair of corresponding interpolation points in the intermediate transition segment. It follows the formula: point_merge = *coef*P1 + (1 - *coef*)*P2. Here, P1 represents the second target interpolation point from the original trajectory, and P2 represents the first target interpolation point from the derived trajectory. When *coef* is 0, the fused point is exactly equal to P2 (the derived trajectory point); as *coef* increases towards 1, the fused point gradually shifts towards P1 (the original trajectory point). This calculation is performed synchronously on all state parameters, including the point's coordinates and heading angle. The technical effect of this step is to achieve a trajectory in the intermediate transition segment that smoothly and continuously transitions from completely following the derived trajectory to completely following the original trajectory.

[0081] Step S1294: Based on the first a fused feature points, the (a+1)th to the Nbth fused feature points, and the last b fused feature points, determine the first fused sub-trajectory.

[0082] Step S1294 is the final trajectory assembly step. It connects the first a fusion feature points determined in step S1291 (entirely from the derived trajectory), the (a+1)th to Nbth fusion feature points (dynamic transition segments) determined in step S1293, and the last b fusion feature points determined in step S1291 (entirely from the original trajectory) in order from the start point to the end point, thus forming a complete and continuous first fusion sub-trajectory. This final trajectory simultaneously possesses three key characteristics: smooth executability at the starting point (automatic steering wheel movement), accuracy at the end point (correct parking posture), and continuity throughout (no state transitions).

[0083] For example, such as Figure 5 As shown, the derived trajectory segment is... Figure 5For the points in part D, L1 is the first original sub-trajectory, L2 is the first derived sub-trajectory, and L3 is the first fused sub-trajectory.

[0084] Regarding step S13, if the original parking trajectory only includes the first segment of the original sub-trajectory, the first fused sub-trajectory is determined as the optimized target parking trajectory for the target vehicle.

[0085] Step S13 defines the output logic of the parking trajectory optimization method in a specific application scenario. This scenario involves the original parking trajectory consisting of only a single segment (the first original sub-trajectory), typically corresponding to a simple parking operation that can be completed with "one turn, no gear shifting required." In this case, since there are no other gear shift points or subsequent sub-trajectories to process, the entire optimization process has already been completed in step S12: by smoothly expanding and merging the starting point of this unique trajectory segment, a first fused sub-trajectory is generated that ensures both a smooth start (steering wheel input) and an accurate endpoint. Therefore, this step directly determines this first fused sub-trajectory as the optimized target parking trajectory for the target vehicle. This demonstrates the completeness and adaptability of the method, automatically adapting the processing flow according to the complexity of the input trajectory, ensuring that an executable and optimized final trajectory is output under any circumstances.

[0086] In summary, the parking trajectory optimization method provided in this application achieves significant technical effects by obtaining an original parking trajectory containing at least one original sub-trajectory and performing a specific first optimization step on the starting end of the first original sub-trajectory. First, by deleting short segments of the original trajectory with high curvature at the starting end and replacing them with derived trajectory segments with lower curvature and longer lengths, the geometric characteristics of the trajectory's starting segment are fundamentally changed. This allows the vehicle to begin moving from a state with a smaller steering wheel angle, effectively solving the technical bottleneck of traditional parking where "turning the steering wheel while stationary" is required to start, achieving a smooth start with "moving the steering wheel," and improving parking experience and efficiency. Second, by aligning the generated initial sub-trajectory through translation and rotation based on the original starting point state, it is ensured that the optimized trajectory starts from the vehicle's actual starting position and posture, guaranteeing the accuracy of the parking starting point and avoiding execution errors caused by trajectory deviation. Furthermore, by fusing the processed derived sub-trajectory with the original sub-trajectory, the generated first fused sub-trajectory adheres to the optimized smoothness characteristics at the beginning and reverts to the accurate endpoint of the original trajectory at the end, thus ensuring smooth continuity throughout while maintaining the positioning accuracy of the parking endpoint. In addition, this method adaptively outputs results by determining the number of trajectory segments. For simple parking scenarios containing only a single trajectory segment, this method can also effectively optimize the starting process, demonstrating the versatility and robustness of the solution. Therefore, this embodiment accurately optimizes the dynamic characteristics of the trajectory starting point without significantly altering the global shape of the original trajectory, balancing starting smoothness, operational naturalness, and parking endpoint accuracy.

[0087] Furthermore, after fusing the first original sub-trajectory and the first derived sub-trajectory to obtain the first fused sub-trajectory, that is, after step S12, if the original parking trajectory includes the first original sub-trajectory and at least one second original sub-trajectory, the method further includes steps S21-S22.

[0088] Step S21: For the first fused sub-trajectory updated from the first original sub-trajectory segment and at least one second original sub-trajectory segment, for every two segments of the sub-trajectory segment with intersection points, including the preceding sub-trajectory and the following sub-trajectory segment, according to the order of the original parking trajectories, perform the second optimization step on each pair of sub-trajectories corresponding to each intersection point to obtain the fused sub-trajectory corresponding to the first fused sub-trajectory segment and at least one second original sub-trajectory segment respectively.

[0089] The second optimization step is explained as follows: Step S211: Perform a first optimization step on the end of the preceding sub-trajectory to obtain a preceding derived trajectory segment. Combine the preceding derived trajectory segment with the trajectory segment retained in the preceding sub-trajectory to obtain a preceding initial sub-trajectory. The preceding sub-trajectory is the target original sub-trajectory in the first optimization step, and the end of the preceding sub-trajectory is the first end in the first optimization step.

[0090] Regarding step S211, its core is to process the end of the preceding sub-trajectory (the preceding sub-trajectory) that constitutes the shift point. This step treats the "end" as the "first end" in the first optimization step and invokes the entire process of the first optimization step. This means that at the end, it deletes a small segment of the original trajectory with a high curvature and extends outward (along the front of the vehicle's direction of travel) to generate a longer derived trajectory segment with a curvature that smoothly decreases to zero. Subsequently, this derived trajectory segment is spliced ​​with the main body of the original trajectory that was not deleted to form the "preceding initial sub-trajectory". The purpose of this step is to independently optimize the end shape of the preceding trajectory so that the trajectory curvature can smoothly transition to zero before reaching the shift point, so that when the vehicle stops and shifts gears, the steering wheel is already straight or at a small angle, preparing for the next dynamic start.

[0091] Step S212: Perform a first optimization step on the starting end of the subsequent sub-trajectory to obtain a subsequent derived trajectory segment. Combine the subsequent derived trajectory segment with the trajectory segment retained in the subsequent sub-trajectory to obtain a subsequent initial sub-trajectory. The subsequent sub-trajectory is the target original sub-trajectory in the first optimization step, and the starting end of the subsequent sub-trajectory is the first end in the first optimization step.

[0092] Regarding step S212, its core is to process the starting end of the subsequent sub-trajectory (the subsequent sub-trajectory) that constitutes the shift point. This step treats the "starting end" as the "first end" in the first optimization step, and similarly calls the first optimization step. Its operation is symmetrical to but opposite in direction to step S211: at this starting end, it deletes a small segment of the original trajectory and extends it in the opposite direction (i.e., the opposite direction in which the vehicle will start to travel) to generate a derived trajectory segment with curvature smoothly increasing from zero, and then splices it with the retained main trajectory to form the "sub-initial sub-trajectory". The purpose of this step is to independently optimize the shape of the starting end of the subsequent trajectory, construct a smooth "starting guide" starting from zero curvature, so that after the vehicle completes the shift, it can start directly from a small steering wheel angle, increasing steering while moving, and achieving a smooth transition of "dynamic shifting". The combination of steps S211 and S212 creates smooth endpoint conditions for the trajectories on both sides of the shift point.

[0093] For example, such as Figure 6 As shown, the intersection of any two adjacent sub-trajectories extends outwards to feature points, corresponding to... Figure 6 The area corresponding to the dotted line within the circle.

[0094] Step S213: Determine an intermediate point based on the front end point of the end of the preceding initial sub-trajectory and the rear end point of the beginning of the following initial sub-trajectory, wherein the intermediate point is located on the line connecting the front end point and the rear end point.

[0095] Regarding step S213, a conventional method for determining intermediate points can be used, such as the geometric midpoint method. This method is simple and direct: ignoring all other factors such as the length and shape of the two trajectories, it treats only the end point of the first trajectory (at the front end) and the beginning point of the second trajectory (at the rear end) as two independent geometric points. The coordinates of the intermediate point are directly taken as the arithmetic mean of these two coordinates, i.e., intermediate point.x = (at the front end.x + at the rear end.x) / 2, intermediate point.y = (at the front end.y + at the rear end.y) / 2. The heading angle of the intermediate point is also taken as the arithmetic mean of the heading angles of the two points (angle wrapping may need to be handled). This method only calculates the average based on the absolute positions and headings of the two endpoints, completely disregarding the length and shape differences between the two trajectory segments. Its potential drawback is that when the lengths of the two trajectory segments differ significantly, the calculated intermediate point may deviate severely from the actual direction of the longer trajectory, leading to unreasonable excessive deformation of the longer trajectory during subsequent alignment operations, thus affecting the smoothness and rationality of the overall trajectory after fusion.

[0096] In addition, step S213 can also be implemented in the following way, which has two different forms but is essentially the same.

[0097] The first method includes: The total length of the sub-track is determined by the sum of the length of the preceding sub-track and the length of the following initial sub-track. The ratio of the length of the preceding track to the total length of the sub-track is determined as the second fusion coefficient. The product of the second fusion coefficient and the trajectory parameters at the rear point is determined as the first data. The difference between 1 and the second fusion coefficient is determined as the third fusion coefficient. The product of the third fusion coefficient and the trajectory parameters at the front point is determined as the second data. The sum of the first data and the second data is determined as the trajectory parameters at the intermediate point.

[0098] The first method dynamically determines the midpoint position by calculating the length ratio of the two trajectory segments. Specifically, it first calculates the total length of the two trajectories, then uses the ratio of the length of the first segment to the total length as a second fusion coefficient. This means that the longer the first segment, the larger this coefficient, and the greater its influence on the midpoint position. When calculating the midpoint, the parameters of the endpoints of the second segment (at the rear endpoint) are weighted according to this coefficient, while the parameters of the endpoints of the first segment (at the front endpoint) are weighted according to (1 - the coefficient), and the sum of the two is the midpoint parameter. The core idea of ​​this design is to adaptively "bias" the midpoint position towards the endpoint of the longer trajectory. This is because the longer trajectory contains more path points, has greater deformation capacity, and can withstand more adjustments; while the shape of the shorter trajectory needs more protection. This method automatically balances the deformation requirements of the two trajectories during docking through the length ratio.

[0099] The second method includes: determining the total length of the sub-track by the sum of the length of the preceding initial sub-track and the length of the following initial sub-track; determining the fourth fusion coefficient by the ratio of the following track length to the total length of the sub-track; determining the third data by the product of the fourth fusion coefficient and the trajectory parameters at the front end point; determining the fifth fusion coefficient by the difference between 1 and the fourth fusion coefficient; determining the fourth data by the product of the fifth fusion coefficient and the trajectory parameters at the rear end point; and determining the trajectory parameters at the intermediate point by the sum of the third data and the fourth data.

[0100] The second method is mathematically identical to the first in both its essence and final effect, differing only in the expression of weight allocation. It first calculates the total length, but then uses the ratio of the length of the latter trajectory segment to the total length as the fourth fusion coefficient. The longer the latter trajectory segment, the larger this coefficient. During calculation, the parameters of the endpoints of the former trajectory segment are weighted according to this coefficient, while the parameters of the endpoints of the latter trajectory segment are weighted according to (1 - coefficient). Although the weighting objects appear interchangeable in the formula compared to the first method, the two coefficients are complementary (i.e., the second fusion coefficient + the fourth fusion coefficient = 1), and the calculation formula is symmetrical, resulting in identical midpoint coordinates and heading angles. This verifies that the two forms are simply two equivalent expressions of the same principle; their core is the weighted average based on the length ratio of the two trajectory segments, thus making the midpoint position more reasonably close to the endpoint of the longer trajectory to optimize the overall alignment.

[0101] like Figure 7 As shown, a is the end point of the initial sub-trajectory at the front end, b is the start point of the initial sub-trajectory at the rear end, and c is the middle point.

[0102] Step S214: Based on the intermediate point, translate and rotate the initial sub-trajectory in front, so that the end point of the initial sub-trajectory in front coincides with the intermediate point, thus obtaining the derived sub-trajectory in front; merge the derived sub-trajectory in front with the initial sub-trajectory in front, thus obtaining the merged sub-trajectory in front.

[0103] Regarding step S214, its core objective is to precisely align the end of the previously optimized trajectory (the initial sub-trajectory) with the calculated midpoint, generating the final usable merged sub-trajectory. This step first performs a rigid body transformation (translation and rotation), moving and rotating the entire initial sub-trajectory so that its original ending point (the leading point) completely coincides with the midpoint in both position and heading angle. This alignment operation ensures that at the shift point, the ending state of the previous trajectory smoothly connects to a unified, optimized midpoint. Subsequently, the aligned trajectory (the derived sub-trajectory) is merged with the original previous sub-trajectory (i.e., the original trajectory before the starting point optimization). This fusion process is similar to that used when processing the starting point, aiming to maintain the original planned shape in the main body of the trajectory while smoothly transitioning its end to the aligned state, thereby ensuring that the vehicle can smoothly straighten the steering wheel and prepare for shifting when approaching the shift point.

[0104] Step S215: Based on the intermediate point, translate and rotate the initial sub-trajectory so that the starting point of the initial sub-trajectory coincides with the intermediate point, thus obtaining the derived sub-trajectory. Merge the derived sub-trajectory with the sub-trajectory to obtain the merged sub-trajectory.

[0105] Regarding step S215, its objective is symmetrical to that of step S214, aiming to process the latter segment of the trajectory to achieve a smooth bidirectional connection at the shift point. This step also uses rigid body transformation to align the starting endpoint (the rear endpoint) of the optimized latter segment of the trajectory (the initial sub-trajectory) to the same midpoint, resulting in the derived sub-trajectory. This ensures that when the vehicle starts after shifting, its starting position and orientation seamlessly connect with the ending state (i.e., the midpoint) before shifting. Next, the aligned derived sub-trajectory is merged with the original rear sub-trajectory. This fusion ensures that the latter segment of the trajectory initially follows the aligned smooth "guide," while the remaining portion reverts to the original planned path, allowing the vehicle to smoothly initiate the next stage of parking maneuvers from the shift point (when the steering wheel is nearing center).

[0106] For details regarding steps S214 and S215, please refer to [link / reference]. Figure 8 'a' is the end point of the initial sub-trajectory, 'b' is the beginning point of the initial sub-trajectory, and 'c' is the midpoint. The two red trajectories are the merged sub-trajectories. Figure 8 (Red line on the right) and the later fused sub-trajectory ( Figure 8 (Red line on the left side of the middle).

[0107] Step S216: If the second original sub-trajectory is not the last segment of the original parking trajectory, the second original sub-trajectory will be used as the preceding sub-trajectory for the next second optimization step.

[0108] Regarding step S216, it defines the iterative processing mechanism in multi-segment trajectory scenarios. This step stipulates that if the currently processed subsequent trajectory segment (the later fused sub-trajectory) is not the last segment of the entire original parking trajectory, then it will be automatically input as the "previous sub-trajectory" into the next "second optimization step" in the processing of the next shift point. This mechanism is crucial because it ensures the chain-like transitivity of shift point processing and the global consistency of the overall trajectory. The smoothing processing of each shift point updates the subsequent trajectory segment it belongs to, and this updated trajectory becomes the starting point for the processing of the next shift point, iterating in this way until all shift points have been processed. This avoids error accumulation and ensures that the entire optimized trajectory generated from beginning to end is continuous, smooth, and self-consistent.

[0109] If the current processed segment of the trajectory (in the post-fusion sub-trajectory) is the last segment of the entire original parking trajectory, then there is no need to repeat the second optimization step.

[0110] Step S22: Determine the parking trajectory formed by the various fused sub-trajectories as the optimized target parking trajectory for the target vehicle.

[0111] Step S22 is the final step in the parking trajectory optimization method, marking the integration and output stage after all segmented optimization processing of the original parking trajectory containing multiple sub-trajectories. The "each fused sub-trajectory" mentioned in this step refers to an independent trajectory segment that has undergone smoothing processing, generated sequentially through the aforementioned steps (such as optimizing the starting point of the first trajectory segment and bidirectional optimization and fusion of each pair of adjacent sub-trajectories at the shift point).

[0112] The core operation of step S22 is to connect these independent fusion sub-trajectories end-to-end according to the order in which the vehicles park, thereby combining and restoring them into a complete, continuous, and globally consistent parking path. This path is then determined as the optimal trajectory for the target vehicle to execute, i.e., the "target parking trajectory." This step ensures that the effects of all local optimizations (dynamic steering wheel start, dynamic smooth gear shifting) are organically combined into a whole, outputting a final solution that balances smoothness, continuity, and terminal accuracy from start to finish.

[0113] In summary, the parking trajectory optimization method provided in this application, for the initial stage of parking, performs curvature smoothing expansion and fusion processing on the starting point of the first trajectory segment, making the curvature of the generated trajectory approach zero near the starting point. This allows the vehicle to start moving from a small steering wheel angle and smoothly increase the steering angle during movement, thus completely eliminating the bottleneck of traditional parking where "turning the steering wheel while stationary" is required to start. This achieves a natural starting method of "turning the steering wheel while moving," significantly improving driver comfort and parking start efficiency. For the gear shifting stage during parking, by independently extending and optimizing the trajectories on both sides of the shift point, calculating the midpoint based on trajectory length weighting, and performing bidirectional alignment and progressive fusion, the steering wheel can smoothly return to center when the vehicle approaches the shift point. After shifting, the vehicle can immediately start following the next trajectory segment from the returned-to-center state, achieving a "silky" dynamic gear shift connection and avoiding the jerking and waiting caused by the vehicle having to significantly turn the steering wheel again before and after gear shifting. In summary, the embodiments of this application, while ensuring the accuracy of the parking endpoint position, achieve smooth steering and dynamic gear shifting throughout the entire process from the starting point to the endpoint, significantly improving the continuity, naturalness of operation, and overall efficiency of the parking process.

[0114] Based on the same inventive concept, the embodiments of this application provide, as follows: Figure 9 The vehicle shown includes: Processor 91; Memory 92 is used to store executable instructions of the processor 91; The processor 91 is configured to execute a parking trajectory optimization method as described above.

[0115] Since the vehicle described in this embodiment is the vehicle used to implement the information processing method in the embodiments of this application, those skilled in the art can understand the specific implementation method and various variations of the vehicle in this embodiment based on the information processing method described in the embodiments of this application. Therefore, how the vehicle implements the method in the embodiments of this application will not be described in detail here. Any vehicle used by those skilled in the art to implement the information processing method in the embodiments of this application falls within the scope of protection of this application.

[0116] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0117] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0118] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0119] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0120] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0121] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A parking trajectory optimization method, characterized in that, The method includes: Obtain the original parking trajectory of the target vehicle at its current position. The original parking trajectory includes at least one original sub-trajectory, which refers to the trajectory segment between the positions of the target vehicle when the target vehicle changes gears at any two adjacent times. A first optimization step is performed on the starting end of the first segment of the original sub-trajectory of the original parking trajectory, where the current position is located, to obtain a first derived trajectory segment. The first segment of the original sub-trajectory is the target original sub-trajectory in the first optimization step, and the starting end of the first segment of the original sub-trajectory is the first end in the first optimization step. The first derived trajectory segment is combined with the trajectory segment retained in the first segment of the original sub-trajectory to obtain a first initial sub-trajectory. Based on the trajectory parameters of the trajectory feature point where the starting end of the first segment of the original sub-trajectory is located before the trajectory segment is deleted, the first initial sub-trajectory is translated and rotated so that the starting point of the first initial sub-trajectory coincides with the trajectory feature point where the starting end of the first segment of the original sub-trajectory is located before the trajectory segment is deleted, to obtain a first derived sub-trajectory. The first segment of the original sub-trajectory and the first derived sub-trajectory are fused to obtain a first fused sub-trajectory. If the original parking trajectory only includes the first segment of the original sub-trajectory, the first fused sub-trajectory is determined as the optimized target parking trajectory of the target vehicle. The first optimization step includes: deleting the original trajectory segment with a first preset length where the first end is located in the target original sub-trajectory, and then adding a derived trajectory segment with a second preset length to the first end of the target original sub-trajectory; the first preset length is less than the second preset length; and the curvature of the derived trajectory segment is less than the curvature of the original trajectory segment.

2. The parking trajectory optimization method as described in claim 1, characterized in that, The step of deleting the original trajectory segment with a first preset length located at the first end of the target original sub-trajectory, and then adding a derived trajectory segment with a second preset length at the first end of the target original sub-trajectory, includes: The original target sub-trajectory is divided into N-1 trajectory units with a preset sub-length. The endpoints of each trajectory unit are denoted as trajectory feature points. The original target sub-trajectory includes N trajectory feature points; N is a positive integer. Starting from the trajectory feature point at the first end of the target original sub-trajectory, delete the original trajectory segment with a first preset length. The original trajectory segment refers to the trajectory segment composed of all trajectory units between the first trajectory feature point and the (M+1)th trajectory feature point, where M is a positive integer and 2 < M < N. Based on the trajectory parameters corresponding to the (M+1)th trajectory feature point, K derived feature points are determined on the side of the (M+1)th trajectory feature point away from the (M+1)th trajectory unit. Among them, there is a preset sub-length distance between each pair of adjacent feature points from the (M+1)th trajectory feature point to the Kth derived feature point. The sum of the straight-line distances between each pair of adjacent feature points from the (M+1)th trajectory feature point to the Kth derived feature point is the second preset length. The difference in steering wheel angles between any two adjacent feature points from the first derived feature point to the (K-1)th derived feature point is the same. The steering wheel angles of each feature point from the (K-1)th derived feature point to the Kth derived feature point are 0. The vehicle heading angles of each feature point from the (K-2)th derived feature point to the Kth derived feature point are the same. K is a positive integer, K>4, and M<K<N. Based on the (M+1)th trajectory feature point to the Kth derived feature point, the derived trajectory segment is obtained.

3. The parking trajectory optimization method as described in claim 2, characterized in that, The trajectory parameters include trajectory coordinates, vehicle heading angle, steering wheel angle, and driving direction; Based on the trajectory parameters corresponding to the (M+1)th trajectory feature point, K derived feature points are determined on the side of the (M+1)th trajectory feature point away from the (M+1)th trajectory unit, including: The steering wheel angle of the (M+1)th trajectory feature point is determined as the steering wheel angle of the first derived feature point; the difference between the steering wheel angles of any two adjacent feature points from the first derived feature point to the (K-1)th derived feature point is the same; the steering wheel angles of the (K-1)th and Kth derived feature points are 0; and the vehicle heading angles of all feature points from the (K-2)th to the Kth derived feature point are the same. Iterate through i from 1 to K-1, and perform the first step for i. The first step includes: Based on the steering wheel angle at the i-th derived feature point and the wheelbase of the target vehicle, the i-th driving radius of the target vehicle at the i-th derived feature point is determined; Based on the preset sub-length between the i-th derived feature point and the (i-1)-th derived feature point and the i-th driving radius, the i-th difference in vehicle heading angle between the i-th derived feature point and the (i-1)-th derived feature point is determined; Based on the vehicle heading angle of the (i-1)th derived feature point and the i-th difference, determine the vehicle heading angle of the i-th derived feature point; Based on the i-th driving radius, the trajectory coordinates of the (i-1)-th derived feature point, and the i-th difference, determine the trajectory coordinates of the i-th derived feature point; After i takes K-1 and performs the first step, the trajectory coordinates of the Kth derivative feature point are determined based on the trajectory coordinates of the K-1th derivative feature point, the vehicle heading angle, and the preset sub-length between the K-1th derivative feature point and the Kth derivative feature point.

4. The parking trajectory optimization method as described in claim 1, characterized in that, The first original sub-trajectory and the first derived sub-trajectory are fused to obtain the first fused sub-trajectory, including: The first segment of the original sub-trajectory is divided into N-1 first trajectory units with a preset sub-length. The endpoints of each first trajectory unit are denoted as first trajectory feature points. Each first trajectory feature point includes a first cumulative length value, which is the sum of the lengths of all first trajectory units before the first trajectory feature point. The first segment of the original sub-trajectory includes N first trajectory feature points; N is a positive integer. The total length of the first derived sub-trajectory is divided into N-1 second trajectory units. The endpoints of each second trajectory unit are called second trajectory feature points. Each second trajectory feature point includes a second cumulative length value, which is the sum of the lengths of all second trajectory units before the second trajectory feature point. The first derived sub-trajectory includes N second trajectory feature points. Using each first trajectory feature point as a reference, determine the first deviation degree of each second trajectory feature point relative to the first trajectory feature point, and perform interpolation processing on each second trajectory feature point based on the first deviation degree to obtain N first target interpolation points; Using each second trajectory feature point as a reference, the second deviation degree of each first trajectory feature point relative to the second trajectory feature point is determined. Based on the second deviation degree, interpolation processing is performed on each first trajectory feature point to obtain N second target interpolation points. The N first target interpolation points and the N second target interpolation points correspond one-to-one along the extension direction of the first original sub-trajectory. Based on N first target interpolation points and N second target interpolation points, N fusion feature points are obtained, and the first fusion sub-trajectory is determined based on the N fusion feature points.

5. The parking trajectory optimization method as described in claim 4, characterized in that, Using each first trajectory feature point as a reference, determine the first deviation degree of each second trajectory feature point relative to the first trajectory feature point. Based on the first deviation degree, perform interpolation processing on each second trajectory feature point to obtain N first target interpolation points, including: Each first trajectory feature point in the first original sub-trajectory is sequentially taken as a first target feature point and the second step is executed to obtain N first target interpolation points corresponding to the first derived sub-trajectory. The second step includes: From the feature points in the first derived sub-trajectory whose second cumulative length value is greater than the first cumulative length of the first target feature point, select the second target feature point with the smallest second cumulative length value; From the feature points in the first derived sub-trajectory whose second cumulative length value is less than the first cumulative length of the first target feature point, select the third target feature point with the largest second cumulative length value; Based on the first cumulative length value of the first target feature point, the second cumulative length value of the second target feature point, and the second cumulative length value of the third target feature point, a first degree of deviation between the second target feature point and the third target feature point is determined. Based on the trajectory parameters of the second target feature point, the trajectory parameters of the third target feature point, and the first deviation degree, the first target interpolation point corresponding to the second target feature point and the third target feature point is determined.

6. The parking trajectory optimization method as described in claim 4, characterized in that, Using each second trajectory feature point as a reference, the second deviation degree of each first trajectory feature point relative to the second trajectory feature point is determined. Based on the second deviation degree, interpolation processing is performed on each first trajectory feature point to obtain N second target interpolation points, including: Each second trajectory feature point in the first derived sub-trajectory is sequentially used as the fourth target feature point and the third step is executed to obtain N second target interpolation points corresponding to the first original sub-trajectory. The third step includes: From the feature points in the first original sub-trajectory segment whose first cumulative length value is greater than the second cumulative length of the fourth target feature point, select the fifth target feature point with the smallest first cumulative length value. From the feature points in the first original sub-trajectory whose first cumulative length value is less than the second cumulative length of the fourth target feature point, select the sixth target feature point with the largest first cumulative length value; Based on the second cumulative length value of the fourth target feature point, the first cumulative length value of the fifth target feature point, and the first cumulative length value of the sixth target feature point, a second degree of deviation between the fifth target feature point and the sixth target feature point is determined. Based on the trajectory parameters of the fifth target feature point, the trajectory parameters of the sixth target feature point, and the second deviation degree, the second target interpolation point corresponding to the fifth target feature point and the sixth target feature point is determined.

7. The parking trajectory optimization method as described in claim 4, characterized in that, Based on N first target interpolation points and N second target interpolation points, N fused feature points are obtained. Based on these N fused feature points, the first fused sub-trajectory is determined, including: The first a interpolation points of the N first target interpolation points are taken as the first a fusion feature points of the first fusion sub-trajectory; the last b interpolation points of the N second target interpolation points are taken as the last b fusion feature points of the first fusion sub-trajectory; a and b are positive integers, a≤8, b≤8; For the (a+1)th to the Nbth first target interpolation points out of N first target interpolation points, and the (a+1)th to the Nbth second target interpolation points out of N second target interpolation points, a first fusion coefficient is determined based on a first formula for each pair of one-to-one corresponding first target interpolation points and second target interpolation points; the first formula is: coef=0.5*{1-cos[(ia)*PI / (Nab)}; Where, coef is the first fusion coefficient of the i-th corresponding first target interpolation point and second target interpolation point among the N first target interpolation points and N second target interpolation points; cos[(ia)*PI / (Nab)] represents the cosine function of (ia)*PI / (Nab); i represents the index of the i-th interpolation point in the extension direction along the first original sub-trajectory from the starting point to the ending point among the N first target interpolation points and N second target interpolation points, i is a positive integer, and a<i<N-b+1; PI represents π; Based on the (a+1)th to the Nbth first target interpolation point among N first target interpolation points, the (a+1)th to the Nbth second target interpolation point among N second target interpolation points, and the first fusion coefficient corresponding to each pair of first target interpolation points and second target interpolation points, the (a+1)th to the Nbth fusion feature points in the first fusion sub-trajectory are determined; Based on the first a fusion feature points, the (a+1)th to the Nbth fusion feature points, and the last b fusion feature points, the first fusion sub-trajectory is determined.

8. The parking trajectory optimization method as described in claim 1, characterized in that, After fusing the first original sub-trajectory and the first derived sub-trajectory to obtain the first fused sub-trajectory, if the original parking trajectory includes the first original sub-trajectory and at least one second original sub-trajectory, the method further includes: For the first fused sub-trajectory updated from the first original sub-trajectory and at least one second original sub-trajectory, each pair of sub-trajectories with intersection points, including the preceding sub-trajectory and the following sub-trajectory, the second optimization step is performed on each pair of sub-trajectories corresponding to each intersection point in the order of the original parking trajectories, to obtain the fused sub-trajectories corresponding to the first fused sub-trajectory and at least one second original sub-trajectory respectively. The parking trajectory formed by the various fused sub-trajectories is determined as the optimized target parking trajectory for the target vehicle. The second optimization step includes: A first optimization step is performed on the end of the preceding sub-trajectory to obtain a preceding derived trajectory segment. The preceding derived trajectory segment is combined with the trajectory segment retained in the preceding sub-trajectory to obtain a preceding initial sub-trajectory. The preceding sub-trajectory is the target original sub-trajectory in the first optimization step, and the end of the preceding sub-trajectory is the first end in the first optimization step. A first optimization step is performed at the starting end of the subsequent sub-trajectory to obtain a subsequent derived trajectory segment. The subsequent derived trajectory segment is combined with the trajectory segment retained in the subsequent sub-trajectory to obtain the subsequent initial sub-trajectory. The subsequent sub-trajectory is the target original sub-trajectory in the first optimization step, and the starting end of the subsequent sub-trajectory is the first end in the first optimization step. An intermediate point is determined based on the front point at the end of the preceding initial sub-track and the rear point at the start of the following initial sub-track, wherein the intermediate point lies on the line connecting the front point and the rear point. Based on the midpoint, translate and rotate the initial sub-trajectory so that the end point of the initial sub-trajectory coincides with the midpoint, thus obtaining the derived sub-trajectory; merge the derived sub-trajectory with the initial sub-trajectory to obtain the merged sub-trajectory. Based on the midpoint, the initial sub-trajectory is translated and rotated so that the starting point of the initial sub-trajectory coincides with the midpoint, resulting in the derived sub-trajectory. The derived sub-trajectory is then merged with the initial sub-trajectory to obtain the merged sub-trajectory. If the second original sub-trajectory is not the last segment of the original parking trajectory, then the second fused sub-trajectory will be used as the preceding sub-trajectory for the next execution of the second optimization step.

9. The parking trajectory optimization method as described in claim 8, characterized in that, The intermediate point is determined based on the point at the end of the preceding initial sub-track and the point at the beginning of the following initial sub-track, including: The total length of the sub-trajectory is determined by the sum of the length of the preceding sub-trajectory and the length of the following sub-trajectory. The ratio of the length of the preceding trajectory to the total length of the sub-trajectories will be determined as the second fusion coefficient; The product of the second fusion coefficient and the trajectory parameters at the back end point is used to determine the first data; The difference between 1 and the second fusion coefficient is determined as the third fusion coefficient; The product of the third fusion coefficient and the trajectory parameters at the front end point is used to determine the second data; The sum of the first and second data points is used to determine the trajectory parameters of the midpoint.

10. A vehicle, characterized in that, include: processor; Memory used to store the processor's executable instructions; The processor is configured to execute a parking trajectory optimization method as described in any one of claims 1 to 9.

Images