Method and system for generating optimal reference path based on road boundary and curvature constraint

Abstract

The application discloses a kind of optimal reference path generation method and system based on real road scene constraint, method includes the following steps: on the basis of lane level navigation result, according to lane center sampling original path point, intercept the road section with curvature less than set value, the remaining path is segmented optimization;Optimize the safety of path based on the driving habits of human, consider the shape of vehicle body, by the preliminary movement of the path point of initial path, the distance difference between vehicle body and the boundary of road two sides does not exceed set value, obtain the path after preliminary optimization;Curve fitting is carried out to the path point on the path after preliminary optimization, optimize the smoothness of path within given minimum error range ε, reduce the number of parameters by the optimization of L0 norm and ensure the smoothness of curvature, generate reference path easy for vehicle tracking and meet the requirements of drivability;Safe, reasonable reference path meeting safety constraint and curvature constraint can be planned.

Description

Technical Field

[0001] This invention belongs to the field of autonomous driving path planning, specifically involving an optimal reference path generation method and system based on road boundary and curvature constraints. Background Technology

[0002] Path planning is one of the core modules for autonomous driving. Global path planning is the upstream module in path planning; it is responsible for finding a feasible path from the current pose to the destination based on environmental information, thus achieving navigation. Furthermore, the global path provides downstream planning modules with a priori reference path for real-time planning and decision-making in dynamic scenarios. Local planning constructs a Frenet coordinate system based on the reference path information and performs sampling-based or optimization-based motion planning. Therefore, the reference path has a significant impact on motion planning, and a reasonable reference path is crucial for motion planning.

[0003] Mainstream methods do not consider vehicle shape when generating reference paths. When traveling along this reference path, the vehicle's rear axle center is controlled to adhere to the path point based on vehicle dynamics. However, for vehicles traveling on structured roads, the vehicle body is not always symmetrical with respect to the road centerline, especially on roads with high curvature. Therefore, paths generated by sampling the road centerline can cause the vehicle to get too close to the boundary, potentially leading to collisions. Global path planning should generate a collision-free, drivable reference path in a static environment. While downstream motion planning can avoid collisions by processing real-time perception results, an inappropriate reference line will inevitably increase the burden on motion planning. This increased burden may even prevent the planning of a feasible path. Conversely, a collision-free reference path offers good performance, at least guaranteeing the existence of a feasible solution and reducing the burden on obstacle avoidance motion planning. Summary of the Invention

[0004] To address the problems existing in the prior art, this invention provides an optimal reference path generation method considering real road scenario constraints. This method can plan a safe and reasonable reference path that satisfies both safety and curvature constraints. The method is based on two stages of optimization. In the first stage, reasonable simplification is introduced to obtain an analytical solution for the distance between the vehicle and the road boundary, establishing a gradient-solvable optimization problem to keep the distance between the vehicle and both sides of the road boundary approximately equal. In the second stage, by solving the L0 norm optimization problem, a spiral-based curve fitting is achieved to improve the feasibility and smoothness of the global path.

[0005] To achieve the above objectives, the technical solution adopted by this invention is: an optimal reference path generation method based on real road scene constraints, comprising the following steps:

[0006] Based on the lane-level navigation results, the original path points are sampled from the lane center, road segments with curvature less than the set value are removed, and the remaining path is segmented and optimized.

[0007] To optimize the safety of the route based on human driving habits, the shape of the vehicle body is taken into account. By initially moving the path points of the initial route, the distance difference between the vehicle body and the two sides of the road does not exceed the set value, and a preliminary optimized route is obtained.

[0008] Curve fitting is performed on the path points on the initially optimized path to optimize the smoothness of the path within a given minimum error range ε. The number of parameters is reduced and the smoothness of the curvature is ensured by optimizing the L0 norm, generating a reference path that is easy for vehicles to follow and meets the requirements of drivability.

[0009] Furthermore, vehicle collisions are considered during the global path generation process, and an objective function is constructed to measure the degree to which the vehicle is centered on the road, thereby optimizing path safety, including:

[0010] The original path points sampled from the center of the lane are used as input. The original path points are moved to optimize safety. The amount of movement of the path points is used as the decision variable to establish the optimization problem.

[0011] Based on a real road scenario, an approximation is introduced to analytically calculate the distance between the vehicle body and the left and right boundaries of the road, and an objective function is constructed accordingly. By reducing the objective function, it is ensured that the difference between the distance between the vehicle body and the left and right boundaries of the road on the path is less than a set value.

[0012] Based on the properties of dense sampling points, the angle and curvature of the path points after movement are analytically calculated, and curvature constraints are incorporated into the safety optimization.

[0013] Furthermore, safety optimization is performed on the movement of the original path points. When establishing the optimization problem with the movement amount of the path points as the decision variable, the path optimization is achieved by moving the path points. The movement amount of the path points in the two-dimensional plane is used as the decision variable, and the decision variable is simplified to a one-dimensional variable. The movement amount of each path point is described by a scalar when the direction is determined.

[0014] Furthermore, when calculating the distance between the vehicle body and the road boundary on both sides, a simplification based on the real road scene is introduced. Specifically, the shape of the local lane centerline from front to back of the vehicle is approximated as an arc, and the road centerline is approximated as a path with constant curvature. Within the vehicle body length range, the road width value is considered constant, and the vehicle body is approximated as a rectangle. Given the lateral offset and orientation of the vehicle body, the distance between the two sides of the rectangular vehicle body and the arc-shaped road centerline is analytically calculated.

[0015] Furthermore, the vehicle orientation is obtained through analytical calculation of the lateral displacement of the path points. In dense sampling points, the orientations of two adjacent points are equal. Through the geometric relationship of triangles, the change in orientation of the path points after movement is obtained, and thus the vehicle orientation after the path points are moved is obtained.

[0016] Furthermore, the curvature calculation is obtained by calculating the change in orientation with mileage.

[0017] Furthermore, the smoothness optimization is achieved through helical segment fitting, so that the optimized path has a piecewise constant rate of curvature change.

[0018] Using the same concept as the method described above, an optimal reference path generation system based on real road scenario constraints is also provided, including a lane sequence acquisition module, a preliminary optimization module, and a global optimization module;

[0019] The lane sequence acquisition module is used to sample the original path points based on the lane center, and cut off road segments with curvature less than a set value, based on the lane-level navigation results.

[0020] The preliminary optimization module is used to optimize the safety of the route based on human driving habits. Taking into account the shape of the vehicle body, it makes preliminary adjustments to the path points of the initial route. The difference in distance between the vehicle body and the two sides of the road does not exceed the set value, thus obtaining a preliminary optimized route.

[0021] The global optimization module is used to perform curve fitting on the path points on the initially optimized path. It optimizes the smoothness of the path within a given minimum error range ε. It reduces the number of parameters and ensures the smoothness of curvature by optimizing the L0 norm, generating a reference path that is easy for vehicles to follow and meets the requirements of drivability.

[0022] Another computer device is provided, including a processor and a memory. The memory is used to store a computer executable program. The processor reads the computer executable program from the memory and executes it. When the processor executes the program, it can realize the optimal reference path generation method based on real road scene constraints described in this invention.

[0023] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can implement the optimal reference path generation method based on real road scenario constraints described in the present invention.

[0024] Compared with the prior art, the present invention has at least the following beneficial effects:

[0025] This invention proposes a simple method for estimating the distance between a vehicle and the road boundary in a realistic road scenario. It analytically calculates the distance based on the vehicle's position and road width. It reveals an approximately linear relationship between the lateral displacement of a path point and its curvature change. The invention proposes a gradient-solvable optimization problem, resulting in an optimized path that positions the vehicle as close to the center of the road as possible, with approximately equal distances from both road boundaries. Combining existing curve fitting methods, the invention provides a method for generating an optimal reference path that considers both path safety and smoothness. This two-stage optimization-based optimal reference path generation method is tested under various road scenarios, verifying the significant optimization and performance improvement of the reference path. Experiments in different scenarios validate the effectiveness of the method: the generated reference path effectively avoids collisions between the vehicle and the road boundary, and possesses better curvature characteristics—piecewise linear curvature and a piecewise constant rate of change of curvature. Therefore, the method of this invention significantly optimizes the safety and smoothness of the reference path. Attached Figure Description

[0026] Figure 1 This is a flowchart for generating the optimal reference path based on road boundaries and curvature constraints.

[0027] Figure 2 This is a diagram illustrating the movement of the initial path point during security optimization.

[0028] Figure 3 This is a schematic diagram for solving the change in orientation of a path point after it moves.

[0029] Figure 4 Distance from both sides of the vehicle body to the center line and A schematic diagram.

[0030] Figure 5 For calculation Schematic diagram

[0031] Figure 6 For calculation Schematic diagram

[0032] Figure 7 To compare experimental results of the road centerline and the optimized path in different road scenarios.

[0033] Figure 8 Two real-world road scenarios for experimental verification

[0034] Figure 9 The results of comparing the curvature of two paths tested in the park's roads.

[0035] Figure 10 Comparison of the curvature of two paths tested in urban roads.

[0036] Figure 11 Comparison of safe distances for two paths tested on park roads.

[0037] Figure 12 Comparison of safe distances for two paths tested in urban roads. Detailed Implementation

[0038] The exemplary embodiments of this application are described in detail below with reference to the accompanying drawings and specific implementations, including various details of the embodiments of this application to aid understanding. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. After reading this invention, any modifications of the invention in various equivalent forms by those skilled in the art fall within the scope defined by the appended claims.

[0039] Figure 1 This is a flowchart illustrating the optimal reference path generation process based on a two-stage optimization method, applicable to various road environments, including extreme scenarios with high curvature and narrow roads. First, the original path points are sampled from the lane center. To reduce the dimensionality of decision variables and improve efficiency, the path is segmented for optimization. The entire optimization process is divided into two stages: In the first stage, this invention optimizes path safety based on human driving habits. The optimized path should be collision-free with road boundaries, and the vehicle should be positioned as close to the lane center as possible. It's important to note that this differs from simply locating a point at the lane center (e.g., the rear axle center of the vehicle). For safety, the shape of the vehicle body must be considered. Therefore, to measure path safety, the distance from the vehicle body to the left and right boundaries should be considered. This stage of optimization is achieved by initially moving the lane centerline, using the amount of movement as the decision variable to establish an optimization problem that meets safety optimization requirements. When path points move, the vehicle's position and orientation on the path change, increasing the difficulty of calculating the shortest distance between the vehicle and the boundaries. This invention introduces reasonable simplifications to give it an analytical solution form, a gradient-solvable optimization model. In the second stage, curve fitting is performed on the path points to optimize the smoothness of the path within a given minimum error range ε. The number of parameters is reduced and the smoothness of the curvature is ensured through L0 norm optimization. Finally, experiments in different road environments verify that the reference path generated by this invention has superior performance. Details of this invention are as follows:

[0040] Step 1: After obtaining the lane-level navigation results, initial path points are generated by sampling at the center of the lane, and sections with low curvature are removed. The remaining path points are then divided into segments for subsequent two-stage optimization. Specific details are as follows:

[0041] First, lane-level navigation is performed to obtain the lane sequence, and the lane center is sampled to generate the initial path point P. initTo make the initial path point P init To accurately reproduce the shape information of the real road, the sampling interval should not be too small; at the same time, to reduce time overhead for subsequent optimization, the sampling interval should not be too large. Experiments have verified that a sampling interval within the range of 1m to 2m can meet these two requirements, ensuring that the sampling points completely reproduce the shape information of the road while satisfying real-time requirements.

[0042] Set curvature threshold κ thre Length threshold L thre On the initial path P init In the middle, remove those with a length greater than L. thre And the curvature remains consistently in [-κ] max ,κ max The path is divided into segments, and the following two-stage optimization is performed on the retained parts.

[0043] Step 2: Construct a safety optimization problem—keeping the distance between the vehicle body and the road boundaries approximately equal. In this step, the input is the initial path P. init The path point after the first stage of optimization is P. prem .

[0044] In this stage, the safety of the path is optimized based on human driving habits. The optimized path should avoid collisions with road boundaries, and the vehicle should be positioned as close to the center of the lane as possible. It's important to note that this differs from simply locating a point at the center of the lane (such as the rear axle center); for safety, the shape of the vehicle body must be considered. Therefore, to measure path safety, the distance from the vehicle body to the left and right boundaries should be considered. This stage of optimization is achieved by adjusting P... init This is accomplished by initially moving the path points, as shown below:

[0045]

[0046]

[0047]

[0048]

[0049] In the above formula, P prem This represents the path after initial optimization at this stage, where D represents P. init The movement of intermediate path points. Therefore, the safety optimization requirement can be expressed as an optimization problem with D as the decision variable:

[0050]

[0051] st-κ max ·1<κ(D)<κ max ·1 (5)

[0052] Where J represents P prem The greater the safety, the larger J becomes. κ(D) consists of the curvature of each path point after motion, and it should satisfy curvature constraints to generate a feasible path for the vehicle to follow. In this stage, we seek a D that keeps the vehicle body in the center of the road, rather than just considering a point on the vehicle. The objective function J involves the calculation of the distance between the vehicle body and the boundary. In optimization problem (5), the objective function includes the shortest distance from the vehicle to the road boundary, which usually cannot be directly calculated by analytical formula, but must be obtained by iterative calculation of the boundary points. Therefore, it is impossible to give an effective differentiable objective function to construct a gradient-based optimization problem. Secondly, calculating the curvature from the position of the point is nonlinear and computationally complex, which increases the computational cost of the optimization problem and leads to non-convex constraints. The solutions to the above problems will be elaborated in detail in steps 3 to 5 respectively.

[0053] Step 3: Based on the above model, establish the relationship between the movement of path points and the change in orientation.

[0054] D represents the motion of the initial path point, and each motion is a vector that includes changes in the X and Y directions in a two-dimensional Cartesian coordinate system. This definition makes subsequent calculations quite complex. Therefore, this invention simplifies the motion model of the point to improve efficiency.

[0055] d (i) It can be broken down into two directions: the tangential direction along the road. and normal direction Point edge The movement of the point mainly changes the distance between the path points, while the point along the path changes the distance between the path points. The motion significantly affects the curvature of the path and the distance of the vehicle from the road boundary, which are precisely the two aspects that this method considers for optimization. Therefore, the path points in this method only move along... Movement, such as Figure 2 As shown, the dark dots represent the initial path P sampled from the road centerline. init The light-colored dots represent the starting point. Initial optimization points after moving a certain distance The positive direction of motion points towards the center of the circle of curvature. yes The orientation of P init It consists of dense sampling points, so it can be accurately estimated from adjacent path points. Therefore, and The relationship between them is as follows:

[0056]

[0057] in yes The curvature, which can be determined by P init The calculations are stored for later use. Therefore, since the direction of motion is determined, the motion of each point can be described by a scalar, and D is simplified to a one-dimensional vector.

[0058] Because of P init Since these are dense points, we can approximate that the directions of two adjacent sampling points are equal, such as... Figure 3 As shown. Therefore, by solving... Figure 3 The right triangle represented by the light-colored lines can be used to calculate Δθ. (i) Their geometric relationships are as follows:

[0059]

[0060] Where L i,i-1 for and The interval between; due to P prem It is a dense point. Angles can be used point to Approximated by the angle. Based on Δθ (i) From the definition of , we can obtain the following relationship:

[0061]

[0062] Therefore, by combining equations (7) and (8), P can be estimated from the properties of dense points. prem The direction.

[0063] Step 4: Calculation method for the distance between the vehicle body and the boundary

[0064] Using scattered information about road boundaries as input for calculation would complicate the problem. Therefore, this invention simplifies the calculation based on real-world road conditions, enabling convenient and accurate calculation of the distance between the vehicle and the boundary. First, to keep the car centered in the lane, the distance from the vehicle to the road boundary should be calculated. This is estimated by the maximum offset of the vehicle body from the road center. Since the distance from the road center to both sides of the boundary is equal, and the road width is easily obtained from high-definition map (HD map) information or perception results, the maximum offset of the vehicle from the road center also reflects the distance between the vehicle and the boundary. If the maximum offsets from the left and right sides of the vehicle to the road center are equal, it can be approximated that the distance between the vehicle and both sides of the boundary is equal, which is the ideal state that this method aims to achieve.

[0065] To simplify the calculation, the shape of the local lane centerline from front to rear of the vehicle is approximated as an arc, such as... Figure 4 As shown. This simplification also reflects actual road conditions, where the road shape is smooth and the curvature does not change drastically in this short section. The vehicle body is approximately rectangular. In this case, the maximum distance between the two sides of the vehicle body and the center of the road can be found, such as... Figure 4 The thick line segment is shown in the image. This indicates the maximum distance between the vehicle and the road centerline on the side closest to the center of the circle of curvature. This represents the distance away from the center of the circle of curvature. Therefore, the distances from both sides of the vehicle to the road boundary can be expressed as follows:

[0066]

[0067] In the formula, This is the distance from the vehicle to the road boundary closest to the center of the circle of curvature. This is the distance from the boundary side that is farther from the center of the circle of curvature. road The width of the road segment is obtained from high-precision map information or real-time sensing. Figure 5 The solution is shown. The method. Equation (10) can be derived from this:

[0068]

[0069] In the formula R (i) for The radius of curvature of the position, w car Let the width be the vehicle width. Combining equations (8) and (10), we can calculate... Figure 6 Demonstrates the solution The method. Using the Law of Cosines, we can obtain the following relationship:

[0070]

[0071] Where L front This refers to the distance from the center of the rear axle to the front of the vehicle, and w car These are also parameters related to the shape of the vehicle body. Combining equations (8), (10), and (11), It can be solved using a specific D.

[0072] Step 5: Construct a specific security optimization problem and clarify the specific form of the objective function and constraints.

[0073] This invention constructs an optimization problem with the movement amount D of the path points as the decision variable, as shown in (5). Based on the conclusions given above, this step will establish the specific form of the optimization problem (5). First, in order to ensure road safety, it is expected that the distance between the vehicle body and the boundaries on both sides of the road is as equal as possible, and the cost function is defined as:

[0074]

[0075] Since formulas (10) and (11) have already been given and The solution method yields the form of the cost function J related to the decision variable D. Then, it is necessary to clarify these constraints. The first stage of optimization needs to roughly ensure that the path points satisfy the curvature constraint, that is, the curvature of the optimization point should be within [-κ]. max ,κ max [In the context of curvature, since curvature is the rate of change of direction with distance, and considering the properties of dense points, the estimate of curvature can be approximated by the following linear relationship:]

[0076]

[0077] Since formula (7) has already established the change in direction Δθ after moving the path point, (i) The relationship with D allows us to obtain P. prem The relationship between curvature and D. The above establishes the optimization problem for the first stage of optimization safety. In addition, the direction calculation can be further simplified to improve efficiency. Considering the actual situation, the direction of the optimized path point will not change significantly. Therefore, the first-order Taylor approximation of (7) can be used to calculate Δθ. (i) :

[0078]

[0079] The constraint becomes d (i) The linear inequalities make the constraints convex, reducing the difficulty of solving the problem.

[0080] Step 6: Smoothness Optimization – Fitting the spiral segment to the path output above

[0081] In this step, P will be fitted with a helical segment. prem Smoothness optimization is performed to ensure the path satisfies kinematic constraints. Paths composed of helical segments have continuous curvature, with piecewise linear curvature consisting of straight lines, arcs, and helices. Such paths have a constant rate of curvature change almost everywhere, effectively reducing path jitter. Lima et al. proposed describing a set of helical fittings to the original scatter points and constraining the error to within ε. This step introduces an L0 norm optimization method during the optimization process. The objective function and inequality constraints are as follows:

[0082] min||W⊙D w κ||1

[0083]

[0084] D here w It is a matrix operator that calculates the second-order difference of a vector, where κ is the piecewise linear curvature vector to be estimated, and x(κ) and y(κ) are the x and y values ​​of the fitted path point coordinates, respectively. raw and y raw Represents the original path P init The coordinates are ε, the maximum allowable error is ⊙, the Hadamard product is ⊙, and W is the weight vector, whose weights are initially set to 1 and updated in each iteration.

[0085]

[0086] Therefore, D w Non-zero values ​​in κ will be forced to be reduced to 0, with a large penalty coefficient. After multiple iterations, D w Only a few non-zero values ​​remain in κ, thus achieving the L0 norm optimization objective.

[0087] The optimal reference path is generated based on two-stage optimization and validated in various different road scenarios.

[0088] Reference paths are generated in various traffic scenarios, and the properties of the generated paths are analyzed. The advantages of this method in terms of safety and smoothness, as well as its applicability, are verified.

[0089] Considering the various road conditions with large curvatures in real traffic scenarios, the following four situations were selected: "C" shaped roundabout, "L" shaped sharp turn, "U" shaped turn and "S" shaped winding road. Figure 7 The images show the vehicle's movement along the road centerline P. init The driving situation (first line) and the optimal reference path P generated by this invention. opti Driving conditions (second line). Figure 7 The method proposed in this invention generates a reference path that keeps the vehicle as far away from the road boundary as possible, thereby avoiding collisions that are unavoidable with the initial path. This significantly improves the safety of the reference path. Furthermore, the table below lists the quantified comparison results. "Safe distance" is defined as the shortest distance between the vehicle and the road boundary, and is negative if a collision occurs. According to the information shown in the table, the path generated by this method significantly improves the safe distance and avoids all collisions. The results indicate that this method can generate reference paths more suitable for conditions with large curvature.

[0090]

[0091]

[0092] Figure 7 The last line also shows P init and P opti The curvature clearly reveals that the latter is not only smoother in curvature, but also results in a smaller curvature when the shape of the road changes, verifying the superiority of the method of the present invention in terms of curvature smoothness.

[0093] This invention has also been verified in real-world scenarios, with satellite images of the selected scenarios shown below. Figure 8 (a) and Figure 8 As shown in (b), the first scenario is a winding road in an industrial park, approximately 140 meters long. The road is almost a straight line with low curvature at its beginning and end, while two adjacent curved sections with opposite curvatures are contained within the road segment, resulting in sections where the curvature changes significantly locally. The second scenario is a simple curved road in an urban setting, including left turns at intersections, approximately 1750 meters long, with relatively low curvature. The optimal reference path P generated using the method proposed in this invention will be... opti The initial path P sampled from the center of the lane init The experimental results for the quantitative comparison of curvature and safety are as follows: Figures 9-12 As shown.

[0094] from Figure 9 and Figure 10 As can be seen, the reference path generated by this method not only has smooth curvature, but also has a small maximum curvature along the path on the curved road, which undoubtedly proves the superiority of this method in terms of smoothness. Figure 11 and Figure 12 This represents the distance difference between the vehicle and the two sides of the boundary, i.e., it measures the degree to which the vehicle is centered on the road. Figure 11 It can be seen that the distance difference on the global path generated in this paper is relatively small, which improves the security of the global path. However, as... Figure 12 As shown, this method increases the distance difference of paths generated along low-curvature roads. This is because the curvature of the road is inherently small (less than 0.01 except for left turns at intersections), so the vehicle body on the road centerline is exactly in the center of the road, meaning the initial path P... init This is already sufficiently safe. In the final stage of our proposed method, spiral segment fitting is performed to optimize smoothness, resulting in a deviation of less than ε in the location of path points. Since ε is defined as a very small given error, it has no significant impact on safety. As shown in the table below, the safe distance of the path generated by this method is only slightly reduced compared to the initial path, but still remains above 0.37.

[0095]

[0096]

[0097] In summary, when the curvature of a road is large, its safety and smoothness are significantly optimized. However, when the curvature of the road is small, the initial path will be safe. In this case, the method proposed in this invention improves the smoothness of the path without significantly affecting safety.

[0098] With the same technical concept as the method, this invention provides an optimal reference path generation system based on real road scenario constraints, including a lane sequence acquisition module, a preliminary optimization module, and a global optimization module;

[0099] The lane sequence acquisition module is used to sample the original path points based on the lane center, and cut off road segments with curvature less than a set value, based on the lane-level navigation results.

[0100] The preliminary optimization module is used to optimize the safety of the route based on human driving habits. Taking into account the shape of the vehicle body, it makes preliminary adjustments to the path points of the initial route. The difference in distance between the vehicle body and the two sides of the road does not exceed the set value, thus obtaining a preliminary optimized route.

[0101] The global optimization module is used to perform curve fitting on the path points on the initially optimized path. It optimizes the smoothness of the path within a given minimum error range ε. It reduces the number of parameters and ensures the smoothness of curvature by optimizing the L0 norm, generating a reference path that is easy for vehicles to follow and meets the requirements of drivability.

[0102] The present invention can also provide a computer device, including a processor and a memory, wherein the memory is used to store a computer executable program, the processor reads the computer executable program from the memory and executes it, and the processor can implement the optimal reference path generation method based on real road scene constraints described in the present invention when executing the computer executable program.

[0103] On the other hand, the present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can implement the optimal reference path generation method based on real road scenario constraints described in the present invention.

[0104] The computer equipment may be a laptop, desktop computer, workstation, or vehicle-mounted computer.

[0105] The processor described in this invention may be a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), or an off-the-shelf programmable gate array (FPGA).

[0106] The memory described in this invention can be an internal storage unit of a laptop, desktop computer, workstation, or vehicle-mounted computer, such as memory or hard disk; or it can be an external storage unit, such as a portable hard disk or flash memory card.

[0107] Computer-readable storage media can include computer storage media and communication media. Computer storage media includes volatile and non-volatile, removable and non-removable media implemented using any method or technology for storing information such as computer-readable instructions, data structures, program modules, or other data. Computer-readable storage media can include: read-only memory (ROM), random access memory (RAM), solid-state drives (SSDs), or optical discs, etc. Random access memory can include resistive random access memory (ReRAM) and dynamic random access memory (DRAM).

[0108] Finally, it should be noted that the above description is only for illustrating specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Those skilled in the art should understand that any modifications or variations made based on the technical solutions and inventive concepts of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for generating an optimal reference path based on road boundary and curvature constraints, characterized in that, Includes the following steps: Based on the lane-level navigation results, the original path points are sampled from the lane center, road segments with curvature less than the set value are removed, and the remaining path is segmented for two-stage optimization. The first stage focuses on safety optimization. Based on human driving habits, it optimizes path safety, considering the vehicle's shape. Initial path points are moved to ensure the distance between the vehicle and the road's side boundaries does not exceed a set value, resulting in a preliminarily optimized path. Specifically, during the global path generation process, vehicle collisions are considered, and an objective function measuring the vehicle's position relative to the road center is constructed to optimize path safety, including: The original path points sampled from the center of the lane are used as input. The original path points are moved to optimize safety. The amount of movement of the path points is used as the decision variable to establish the optimization problem. Based on a real road scenario, an approximation is introduced to analytically calculate the distance between the vehicle body and the left and right boundaries of the road, and an objective function is constructed accordingly. By reducing the objective function, it is ensured that the difference between the distance between the vehicle body and the left and right boundaries of the road on the path is less than a set value. Based on the properties of dense sampling points, the angle and curvature of the path points after movement are analytically calculated, and curvature constraints are added to the safety optimization; the optimized path has no collision with the road boundary, and the vehicle is located in the center of the lane as much as possible; The second stage involves smoothness optimization. The path points on the initially optimized path are fitted with spiral segments to optimize the smoothness of the path within a given minimum error range ε. During the optimization process, the number of parameters is reduced and the smoothness of curvature is ensured through optimization of the L0 norm, so that the optimized path has a piecewise constant rate of curvature change. A reference path is generated based on two-stage optimization.

2. The optimal reference path generation method based on road boundary and curvature constraints according to claim 1, characterized in that, When optimizing the safety of the original path point movement, the optimization problem is established by using the movement amount of the path point as the decision variable. The path optimization is achieved by moving the path point. The movement amount of the path point in the two-dimensional plane is used as the decision variable, and the decision variable is simplified to a one-dimensional variable. The movement amount of each path point is described by a scalar when the direction is determined.

3. The optimal reference path generation method based on road boundary and curvature constraints according to claim 1, characterized in that, When calculating the distance between the vehicle body and the road boundary on both sides, a simplification based on the real road scene is introduced. Specifically, the shape of the local lane centerline from front to back of the vehicle is approximated as an arc, and the road centerline is approximated as a path with constant curvature. Within the vehicle body length range, the road width is considered constant, and the vehicle body is approximated as a rectangle. Given the lateral offset and orientation of the vehicle body, the distance from the two sides of the rectangular vehicle body to the arc-shaped road centerline is analytically calculated.

4. The optimal reference path generation method based on road boundary and curvature constraints according to claim 3, characterized in that, The vehicle orientation is obtained by analytical calculation of the lateral displacement of the path points. In dense sampling points, the orientation of two adjacent points is equal. Through the geometric relationship of triangles, the change in orientation of the path points after movement is obtained, and thus the vehicle orientation after the path points are moved is obtained.

5. The optimal reference path generation method based on road boundary and curvature constraints according to claim 1, characterized in that, The curvature calculation is obtained by calculating the change in orientation with mileage.

6. A system for generating an optimal reference path based on road boundary and curvature constraints as described in claim 1, characterized in that, It includes a lane sequence acquisition module, a preliminary optimization module, and a global optimization module; The lane sequence acquisition module is used to sample the original path points based on the lane center, and cut off road segments with curvature less than a set value, based on the lane-level navigation results. The preliminary optimization module is used to optimize the safety of the route based on human driving habits. Taking into account the shape of the vehicle body, it makes preliminary adjustments to the path points of the initial route. The difference in distance between the vehicle body and the two sides of the road does not exceed the set value, thus obtaining a preliminary optimized route. The global optimization module is used to perform curve fitting on the path points on the initially optimized path. It optimizes the smoothness of the path within a given minimum error range ε, reduces the number of parameters and ensures the smoothness of curvature by optimizing the L0 norm, and generates a reference path.

7. A computer device, characterized in that, It includes a processor and a memory, the memory being used to store a computer-executable program, the processor reading part or all of the computer-executable program from the memory and executing it, and the processor executing part or all of the computed executable program is able to implement the optimal reference path generation method based on road boundary and curvature constraints as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, A computer-readable storage medium stores a computer program that, when executed by a processor, enables the optimal reference path generation method based on road boundary and curvature constraints as described in any one of claims 1 to 5.

Images