A fast solution optimization method and system for occupying grid trajectory planning

Abstract

This application relates to the field of autonomous driving technology, specifically a fast solution optimization method and system for occupancy grid trajectory planning. This application combines occupancy grid downsampling and a low-resolution occupancy grid spatial index structure, initial trajectory planning based on fitted curves, and local fine-grained path planning based on trust regions. Through triple optimization and synergistic effects, the solution speed is improved by 8-10 times in actual tests, making real-time planning based on occupancy grids possible. The dynamic basis functions ensure that the generated trajectory itself conforms to vehicle kinematics, improving trajectory quality and executability. The sparsity perception and hierarchical optimization strategies fully utilize the structural characteristics of the problem, avoiding unnecessary computation and achieving precise allocation of computational resources.

Description

Technical Field

[0001] This application relates to the field of autonomous driving technology, specifically a fast solution optimization method and system for occupying grid trajectory planning. Background Technology

[0002] In autonomous driving, an occupancy grid is a method of representing the environment as three-dimensional voxels, where each voxel stores the probability of being occupied by an obstacle.

[0003] When the occupied grid is used directly for trajectory planning, the cost function of the planning problem typically includes an obstacle distance term that is integrated over the occupied grid. This results in a huge computational overhead, requiring the traversal of a large number of voxels in the high-dimensional grid for each cost function evaluation.

[0004] Traditional nonlinear optimization solvers (such as SQP and IPOPT) converge slowly on such problems, failing to meet the stringent requirements of real-time planning for autonomous driving (which must be completed within 100 milliseconds). Computational bottlenecks have become a key obstacle hindering the practical application of grid-based real-time planning. Summary of the Invention

[0005] In view of this, the purpose of this application is to provide a fast solution optimization method and system for occupying grid trajectory planning, so as to solve the problems in the background art.

[0006] To achieve the above objectives, this application adopts the following technical solution:

[0007] This application provides a fast solution optimization method for occupying grid trajectory planning, comprising the following steps:

[0008] Obtain the original 3D occupancy mesh output by the occupancy mesh model, wherein the original 3D occupancy mesh includes multiple original meshes and the occupancy probability of the original meshes;

[0009] Non-empty grids are selected from the original 3D occupied grid, and the non-empty grids are downsampled to obtain a low-resolution occupied grid; and a spatial index structure of the low-resolution occupied grid is constructed.

[0010] Based on the spatial index structure, fast trajectory optimization is performed on the low-resolution occupied grid to obtain an initial trajectory, wherein the initial trajectory is generated based on a fitted curve;

[0011] Based on the initial trajectory, a trust domain is divided in the original 3D occupied grid, and the initial trajectory is finely optimized in the trust domain to obtain the target trajectory.

[0012] In one embodiment of this application, non-empty grids are selected from the original 3D occupancy grid, and the non-empty grids are downsampled to obtain a low-resolution occupancy grid, including:

[0013] The occupancy probability of each original cell in the original 3D occupancy grid is compared with a preset probability threshold, and the original voxels with an occupancy probability greater than the probability threshold are designated as non-empty cells, thus obtaining a set of non-empty cells. The non-empty grid set The mathematical expression is:

[0014]

[0015] In the formula, Represents the coordinates of the original grid. This represents the probability that the original grid is occupied. Indicates the probability threshold;

[0016] The non-empty grid is downsampled to 1 / 4 or 1 / 8 resolution to obtain a low-resolution occupied grid.

[0017] In one embodiment of this application, constructing the spatial index structure of the low-resolution occupancy grid includes:

[0018] The spatial index structure of the low-resolution occupied grid is constructed based on the KDtree algorithm.

[0019] In one embodiment of this application, fast trajectory optimization is performed on the low-resolution occupied grid based on the spatial index structure to obtain an initial trajectory, including:

[0020] Choose the basis function of the fitted curve and based on the basis functions The vehicle trajectory is parameterized as a linear combination of a set of basis functions, resulting in a linear combination trajectory. Among them, linear combination trajectory The mathematical expression is:

[0021]

[0022] In the formula, is the index of the basis function. For a point in time, Time point Location, These are the coefficients of the basis functions. Let be the basis function row vector. For coefficient vectors;

[0023] Based on the linear combination trajectory Construct a coarse optimization problem, wherein the coarse optimization problem includes a smoothing cost and an obstacle penalty;

[0024] The coarse optimization problem is optimized using an optimizer to obtain the initial trajectory. .

[0025] In one embodiment of this application, the mathematical expression of the coarse optimization problem is:

[0026]

[0027]

[0028]

[0029] In the formula, Indicates the cost of combination. Represents the cost of smoothness. This indicates the weight of the obstacle penalty. Indicates a rough obstacle penalty. The matrix representing the transpose of the coefficient vector. Indicates the planning duration. The row vector representing the second derivatives of the basis functions. The transpose of the row vectors representing the second derivatives of the basis functions. Indicates the number of trajectory sampling points. Indicates the weight of the sampling points. This indicates a non-empty grid retrieved through the spatial index structure of the low-resolution occupied grid. This indicates the probability of a low-resolution grid cell being occupied. Represents the distance decay function. This indicates the distance from the trajectory point to the low-resolution grid cell.

[0030] In one embodiment of this application, dividing a trust domain in the original 3D occupancy mesh based on the initial trajectory includes:

[0031] Based on the initial trajectory, the region that is less than a preset distance threshold from the initial trajectory is designated as the trust region.

[0032] In one embodiment of this application, fine-tuning the initial trajectory within the trust domain to obtain the target trajectory includes:

[0033] Extract the non-empty grids within the trust domain and construct the local spatial index structure of the non-empty grids within the trust domain;

[0034] A refined optimization problem is constructed based on the local spatial index, wherein the mathematical expression of the refined optimization problem is:

[0035]

[0036]

[0037]

[0038] In the formula, Indicates the cost of combination. Represents the cost of smoothness. This indicates the weight of the obstacle penalty. This indicates penalties for obstacles within the trust domain. The matrix representing the transpose of the coefficient vector. Indicates the planning duration. The row vector representing the second derivatives of the basis functions. The transpose of the row vectors representing the second derivatives of the basis functions. Indicates the number of trajectory sampling points. Indicates the weight of the sampling points. This represents a non-empty grid retrieved through the local spatial index structure. This represents the probability of an originally occupied grid being occupied. Represents the distance decay function. This represents the distance from the trajectory point to the original occupied grid cell;

[0039] The optimization problem is solved using an optimizer to obtain the target trajectory.

[0040] This application also provides a fast solution optimization system for occupying grid trajectory planning, including:

[0041] The acquisition module is used to acquire the original 3D occupancy grid output by the occupancy grid model, wherein the original 3D occupancy grid includes multiple original grids and the occupancy probability of the original grids;

[0042] The preprocessing module is used to filter out non-empty grids from the original 3D occupancy grid, downsample the non-empty grids to obtain a low-resolution occupancy grid, and construct the spatial index structure of the low-resolution occupancy grid.

[0043] A coarse planning module is used to perform fast trajectory optimization on the low-resolution occupied grid based on the spatial index structure to obtain an initial trajectory, wherein the initial trajectory is generated based on a fitted curve;

[0044] The fine planning module is used to divide the initial trajectory into trust domains in the original 3D occupied grid based on the initial trajectory, and to finely optimize the initial trajectory in the trust domains to obtain the target trajectory.

[0045] This application also provides an electronic device, including: a processor and a memory;

[0046] The memory is used to store computer programs, and the processor is used to execute the computer programs stored in the memory to cause the electronic device to perform the methods described above.

[0047] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described above.

[0048] The beneficial effects of this application are as follows: This application provides a fast solution optimization method and system for occupied grid trajectory planning. This application combines occupied grid downsampling and a low-resolution occupied grid spatial index structure, initial trajectory planning based on fitted curves, and local fine-grained path planning based on trust regions. Through triple optimization and synergistic effects, the solution speed is improved by 8-10 times in actual tests, making real-time planning based on occupied grids possible. The dynamic basis functions ensure that the generated trajectory itself conforms to vehicle kinematics, improving the quality and executability of the trajectory. The sparsity awareness and hierarchical optimization strategies fully utilize the structural characteristics of the problem, avoid unnecessary computation, and achieve precise allocation of computational resources. Attached Figure Description

[0049] The present application will be further described below with reference to the accompanying drawings and embodiments:

[0050] Figure 1 This is a flowchart illustrating a fast solution optimization method for occupying grid trajectory planning in one embodiment of this application;

[0051] Figure 2 This is a flowchart illustrating a specific implementation of a fast solution optimization method for occupying grid trajectory planning in one embodiment of this application;

[0052] Figure 3 This is a structural diagram of a fast solution optimization system for occupying grid trajectory planning, as shown in one embodiment of this application. Detailed Implementation

[0053] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0054] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. Therefore, the drawings only show the layers related to this application and are not drawn according to the actual number, shape and size of the layers in the actual implementation. In the actual implementation, the form, number and proportion of each layer can be arbitrarily changed, and the layer layout may also be more complex.

[0055] Numerous details are explored in the following description to provide a more thorough explanation of embodiments of this application; however, it will be apparent to those skilled in the art that embodiments of this application may be practiced without these specific details.

[0056] Figure 1 This is a flowchart illustrating a fast solution optimization method for occupying grid trajectory planning in one embodiment of this application, as shown below. Figure 1 As shown in this embodiment, a fast solution optimization method for occupancy grid trajectory planning includes the following steps:

[0057] S110, Obtain the original 3D occupancy grid output by the occupancy grid model, wherein the original 3D occupancy grid includes multiple original grids and the occupancy probability of the original grids;

[0058] S120, non-empty grids are selected from the original 3D occupancy grid, and the non-empty grids are downsampled to obtain a low-resolution occupancy grid; and a spatial index structure of the low-resolution occupancy grid is constructed.

[0059] S130, Based on the spatial index structure, perform fast trajectory optimization on the low-resolution occupied grid to obtain an initial trajectory, wherein the initial trajectory is generated based on a fitted curve;

[0060] S140, using the initial trajectory as a reference, divide the trust domain in the original 3D occupied grid, and perform fine optimization on the initial trajectory in the trust domain to obtain the target trajectory.

[0061] This application generally employs three strategies to achieve rapid solution for trajectory optimization: Strategy 1: Dimensionality reduction based on basis functions with dynamic constraints: The trajectory to be planned is parameterized as a linear combination of a set of predefined basis functions, wherein the basis functions themselves satisfy the vehicle's dynamic subspace constraints (such as curvature continuity and bounded acceleration). By optimizing the coefficients of the basis functions rather than all points on the trajectory, the high-dimensional optimization problem is projected into a low-dimensional parameter space, greatly reducing the search range.

[0062] Strategy 2: Accelerating computation through sparsity awareness: Leveraging the sparsity of the 3D occupancy grid in space, when calculating the obstacle distance cost, only non-empty voxels with an occupancy probability higher than the noise threshold are computed. By constructing a spatial index structure for non-empty voxels (such as a KD-Tree or Octree), global dense traversal is transformed into efficient local queries.

[0063] Strategy 3: Hierarchical Refinement Optimization Strategy: The solution process is divided into two stages:

[0064] Coarse optimization stage: On the downsampled low-resolution version of the occupied grid, an initial coarse trajectory is obtained.

[0065] Fine optimization stage: Using the coarse trajectory as the initial solution, a spatial trust domain is defined around it, and fine trajectory optimization is performed only on the original high-resolution occupied grid within the trust domain.

[0066] Figure 2 This is a flowchart illustrating a specific implementation of a fast solution optimization method for occupancy grid trajectory planning in one embodiment of this application. Figure 2 As shown, the implementation process of this application includes:

[0067] (1) Downsampling the 3D mesh and constructing a spatial index structure for the low-resolution mesh;

[0068] After receiving the 3D occupancy mesh output from the vehicle interior occupancy mesh model, this application first filters the original occupancy mesh to construct a non-empty mesh set in order to improve the efficiency of subsequent processing. Then, downsampling is performed to generate a low-resolution occupancy mesh, specifically including:

[0069] (1-1) Compare the occupied probability of each original cell in the original 3D occupied grid with a preset probability threshold, and take the original voxels whose occupied probability is greater than the probability threshold as non-empty cells to obtain a non-empty cell set. The non-empty grid set The mathematical expression is:

[0070]

[0071] In the formula, Represents the coordinates of the original grid. This represents the probability that the original grid is occupied. Indicates the probability threshold;

[0072] (1-2) The non-empty grid is downsampled to 1 / 4 or 1 / 8 resolution to obtain a low-resolution occupied grid.

[0073] In real-world road scenarios, obstacles (vehicles, pedestrians, curbs) occupy only 1% to 10% of the total space (for example, in a 200×200×50 grid, obstacle voxels account for approximately 5%). Therefore, by filtering, most invalid voxels can be eliminated, reducing the query volume in subsequent processing.

[0074] Furthermore, further downsampling can reduce resolution and the computational burden of subsequent processing. Coarse path planning can be performed quickly in a low-resolution grid, and then a trust domain can be generated after obtaining the coarse path plan. Fine optimization can be performed within the trust domain, which can greatly improve the overall planning efficiency.

[0075] After constructing the low-resolution occupancy grid, a spatial index structure is also needed. This is because calculating obstacle distances is the most computationally intensive part of occupancy grid trajectory planning. By constructing a KDtree-based spatial index structure, the computational complexity of traversing voxels can be effectively reduced.

[0076] (2) Problem dimensionality reduction

[0077] This application selects basis functions for the fitted curve (such as a B-spline curve or a polynomial). and based on the basis functions The vehicle trajectory is parameterized as a linear combination of a set of basis functions, resulting in a linear combination trajectory. Among them, linear combination trajectory The mathematical expression is:

[0078]

[0079]

[0080]

[0081] In the formula, is the index of the basis function. For a point in time, Time point Location, These are the coefficients of the basis functions. Let be the basis function row vector. For coefficient vectors;

[0082] In this application, polynomials or B-splines are selected as basis functions based on the vehicle dynamics characteristics, representing the trajectory as a combination of a few coefficients. This reduces the dimensionality of the problem and the computational load.

[0083] (3) Rough planning

[0084] The original occupied grid is downsampled to 1 / 4 or 1 / 8 resolution. On this low-resolution grid, a rough, collision-free path is quickly solved using any nonlinear optimizer. The process includes:

[0085] (3-1) Based on the linear combination trajectory Construct a coarse optimization problem, wherein the coarse optimization problem includes a smoothing cost and an obstacle penalty;

[0086] The mathematical expression for the rough optimization problem is:

[0087]

[0088]

[0089]

[0090] In the formula, Indicates the cost of combination. Represents the cost of smoothness. This indicates the weight of the obstacle penalty. Indicates a rough obstacle penalty. The matrix representing the transpose of the coefficient vector. Indicates the planning duration. The row vector representing the second derivatives of the basis functions. The transpose of the row vectors representing the second derivatives of the basis functions. Indicates the number of trajectory sampling points. Indicates the weight of the sampling points. This indicates a non-empty grid retrieved through the spatial index structure of the low-resolution occupied grid. This indicates the probability of a low-resolution grid cell being occupied. Represents the distance decay function. This indicates the distance from the trajectory point to the low-resolution grid cell.

[0091] Specifically, the decay function The mathematical expression is:

[0092]

[0093] In the formula, Representing distance, equivalent to , This represents the safe distance parameter, which is generally taken as 1.0-2.0 meters. The larger the value, the smaller the impact of obstacles.

[0094] In the above formula, the smoothness cost is essentially the square integral of the acceleration change (avoiding sharp turns). Since obstacles are highly sparse in 3D space, the penalty is calculated using non-empty voxels queried via a KD-Tree query. This application sets a distance decay function to calculate the penalty value. And through... Weighted averaging yields the optimal solution between computational efficiency and security accuracy.

[0095] (3-2) The coarse optimization problem is optimized using an optimizer to obtain the initial trajectory. .

[0096] Finally, this application uses the L-BFGS optimizer to solve the coarse optimization problem and obtains the initial trajectory. .

[0097] (4) Detailed planning

[0098] Fine-tuning uses the trajectory obtained from coarse optimization as the initial value for high-resolution optimization. A trust region is defined around this trajectory (e.g., 1 meter on each side of the trajectory point). During the fine-tuning phase, when calculating the cost function, only the high-resolution occupied grid information within the trust region is queried. This avoids unnecessary global calculations, specifically including:

[0099] (4-1) Based on the initial trajectory, the region that is less than a preset distance threshold from the initial trajectory is taken as the trust domain.

[0100] The mathematical expression for the trust domain is:

[0101]

[0102] In the formula, For position vectors, This is the distance threshold.

[0103] (4-2) Extract the non-empty grids within the trust domain and construct the local spatial index structure of the non-empty grids within the trust domain;

[0104] (4-3) Construct a refined optimization problem based on the local spatial index, wherein the mathematical expression of the refined optimization problem is:

[0105]

[0106]

[0107]

[0108] In the formula, Indicates the cost of combination. Represents the cost of smoothness. This indicates the weight of the obstacle penalty. This indicates penalties for obstacles within the trust domain. The matrix representing the transpose of the coefficient vector. Indicates the planning duration. The row vector representing the second derivatives of the basis functions. The transpose of the row vectors representing the second derivatives of the basis functions. Indicates the number of trajectory sampling points. Indicates the weight of the sampling points. This represents a non-empty grid retrieved through the local spatial index structure. This represents the probability of an originally occupied grid being occupied. Represents the distance decay function. This represents the distance from the trajectory point to the original occupied grid cell;

[0109] (4-4) Use the optimizer to solve the fine optimization problem and obtain the target trajectory.

[0110] The optimization problem and solution process are similar to those described above; please refer to the previous text for understanding.

[0111] The main features and beneficial effects of this application include:

[0112] Significant speed improvement: Through triple optimization and synergistic effect, the solution speed has been increased by 8-10 times in actual tests, making real-time planning based on occupied grid possible.

[0113] Ensuring feasibility: The basis functions based on dynamics ensure that the generated trajectory itself conforms to vehicle kinematics, improving the quality and executability of the trajectory.

[0114] Resource efficiency: The sparsity awareness and hierarchical optimization strategies make full use of the structural features of the problem, avoid unnecessary computation, and achieve precise allocation of computing resources.

[0115] like Figure 3 As shown, this application also provides a fast solution optimization system for occupancy grid trajectory planning, comprising:

[0116] The acquisition module is used to acquire the original 3D occupancy grid output by the occupancy grid model, wherein the original 3D occupancy grid includes multiple original grids and the occupancy probability of the original grids;

[0117] The preprocessing module is used to filter out non-empty grids from the original 3D occupancy grid, downsample the non-empty grids to obtain a low-resolution occupancy grid, and construct the spatial index structure of the low-resolution occupancy grid.

[0118] A coarse planning module is used to perform fast trajectory optimization on the low-resolution occupied grid based on the spatial index structure to obtain an initial trajectory, wherein the initial trajectory is generated based on a fitted curve;

[0119] The fine planning module is used to divide the initial trajectory into trust domains in the original 3D occupied grid based on the initial trajectory, and to finely optimize the initial trajectory in the trust domains to obtain the target trajectory.

[0120] This application presents a fast solution optimization method and system for occupancy grid trajectory planning. This method combines occupancy grid downsampling and a low-resolution occupancy grid spatial index structure, initial trajectory planning based on fitted curves, and local fine-grained path planning based on trust regions. Through triple optimization and synergistic effects, the solution speed is improved by 8-10 times in practical tests, enabling real-time planning based on occupancy grids. The dynamic basis functions ensure that the generated trajectory conforms to vehicle kinematics, improving trajectory quality and executability. The sparsity awareness and hierarchical optimization strategies fully utilize the structural characteristics of the problem, avoiding unnecessary computation and achieving precise allocation of computational resources.

[0121] This embodiment also provides an electronic terminal, including: a processor and a memory;

[0122] The memory is used to store computer programs, and the processor is used to execute the computer programs stored in the memory so that the terminal performs any of the methods in this embodiment.

[0123] As will be understood by those skilled in the art, the computer-readable storage medium described in this embodiment allows for the implementation of all or part of the steps in the above method embodiments by computer program-related hardware. The aforementioned computer program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0124] The electronic terminal provided in this embodiment includes a processor, a memory, a transceiver, and a communication interface. The memory and the communication interface are connected to the processor and the transceiver and complete communication between them. The memory is used to store computer programs, the communication interface is used to perform communication, and the processor and the transceiver are used to run the computer programs, so that the electronic terminal performs the various steps of the above method.

[0125] In this embodiment, the memory may include random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device.

[0126] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0127] In the above embodiments, although the present application has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art based on the foregoing description. The embodiments of the present application are intended to cover all such substitutions, modifications, and variations falling within the broad scope of the appended claims.

[0128] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.

Claims

1. A fast solution optimization method for occupancy grid trajectory planning, characterized in that, The method comprises the steps of: obtaining an original 3D occupancy grid output by an occupancy grid model, wherein the original 3D occupancy grid comprises a plurality of original grids and an occupied probability of the original grids; screening non-empty grids from the original 3D occupancy grid, and performing down-sampling on the non-empty grids to obtain a low-resolution occupancy grid; and constructing a spatial index structure of the low-resolution occupancy grid; performing fast trajectory optimization in the low-resolution occupancy grid based on the spatial index structure to obtain an initial trajectory, wherein the initial trajectory is generated based on a fitting curve; dividing a trust region in the original 3D occupancy grid based on the initial trajectory, and performing fine optimization on the initial trajectory in the trust region to obtain a target trajectory.

2. The fast solution optimization method for occupancy grid trajectory planning according to claim 1, wherein, screening non-empty grids from the original 3D occupancy grid, and performing down-sampling on the non-empty grids to obtain a low-resolution occupancy grid, comprising: The occupied probability of each original grid in the original 3D occupancy grid is compared with a preset probability threshold, and an original voxel with an occupied probability greater than the probability threshold is taken as a non-empty grid to obtain a non-empty grid set Wherein, the non-empty grid set The mathematical expression of the non-empty grid set is: wherein, denotes the coordinates of the original grid, denotes the occupancy probability of the original grid, denotes a probability threshold; performing down-sampling on the non-empty grids to 1 / 4 or 1 / 8 resolution to obtain a low-resolution occupancy grid.

3. The fast solution optimization method for occupancy grid trajectory planning of claim 1, wherein, constructing a spatial index structure of the low-resolution occupancy grid, comprising: constructing the spatial index structure of the low-resolution occupancy grid based on a KDtree algorithm.

4. The fast solution optimization method for occupancy grid trajectory planning of claim 1, wherein, performing fast trajectory optimization in the low-resolution occupancy grid based on the spatial index structure to obtain an initial trajectory, comprising: Selecting basis functions for curve fitting and parameterizing the vehicle trajectory as a linear combination of a set of basis functions, resulting in a linear combination trajectory where the linear combination trajectory is mathematically expressed as:​ wherein is the index of the basis function, is the time point, time point the position of, is the basis function coefficient, is the basis function row vector, is the coefficient vector; based on the linear combination trajectory constructing a coarse optimization problem, wherein the coarse optimization problem includes a smooth cost and an obstacle penalty; optimizing the coarse optimization problem using an optimizer to obtain an initial trajectory .

5. The fast solution optimization method for occupancy grid trajectory planning according to claim 4, wherein, a mathematical expression of the coarse optimization problem is: wherein denotes a combination cost, denotes a smoothness cost, denotes a weight of an obstacle penalty, denotes a coarse obstacle penalty, denotes a transpose matrix of a coefficient vector, denotes a planning horizon, denotes a row vector of a second derivative of a basis function, denotes a transpose matrix of a row vector of a second derivative of a basis function, denotes a number of trajectory sampling points, denotes a sampling point weight, denotes a non-empty grid queried by a spatial index structure of the low resolution occupancy grid, denotes an occupancy probability of a low resolution occupancy grid, denotes a distance decay function, denotes a distance of a trajectory point to a low resolution occupancy grid.

6. The fast solution optimization method for occupancy grid trajectory planning of claim 1, wherein, dividing a trust region in the original 3D occupancy grid based on the initial trajectory, comprising: dividing a region less than a preset distance threshold from the initial trajectory as the trust region based on the initial trajectory.

7. The method of claim 4, wherein, performing fine optimization on the initial trajectory in the trust region to obtain a target trajectory, comprising: extracting non-empty grids in the trust region, and constructing a local spatial index structure of the non-empty grids in the trust region; constructing a fine optimization problem based on the local spatial index structure, wherein a mathematical expression of the fine optimization problem is: In the formula, denotes a combination cost, denotes a smoothness cost, denotes a weight of an obstacle penalty, denotes an obstacle penalty in a trust region, denotes a transposed matrix of a coefficient vector, denotes a planning duration, denotes a row vector of a second derivative of a basis function, denotes a transposed matrix of a row vector of a second derivative of a basis function, denotes a number of trajectory sampling points, denotes a sampling point weight, denotes a non-empty grid queried through the local space index structure, denotes an occupied probability of an original occupancy grid, denotes a distance decay function, denotes a distance of a trajectory point to an original occupancy grid; solving the fine optimization problem using an optimizer to obtain the target trajectory.

8. A fast solution optimization system for occupancy grid trajectory planning, characterized in that, comprising: an acquisition module configured to acquire an original 3D occupancy grid output by an occupancy grid model, wherein the original 3D occupancy grid comprises a plurality of original grids and an occupied probability of the original grids; a preprocessing module configured to screen non-empty grids from the original 3D occupancy grid, and perform down-sampling on the non-empty grids to obtain a low-resolution occupancy grid; and construct a spatial index structure of the low-resolution occupancy grid; a coarse planning module configured to perform fast trajectory optimization in the low-resolution occupancy grid based on the spatial index structure to obtain an initial trajectory, wherein the initial trajectory is generated based on a fitting curve; a fine planning module configured to divide a trust region in the original 3D occupancy grid based on the initial trajectory, and perform fine optimization on the initial trajectory in the trust region to obtain a target trajectory.

9. An electronic device, comprising: comprising: a processor and a memory; the memory is configured to store a computer program, and the processor is configured to execute the computer program stored in the memory, so that the electronic device performs the method in any one of claims 1 to 7.

10. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program, which is executed by a processor, implements the method as claimed in any one of claims 1 to 7.

Images