Self-adaptive unrouted path planning system and method based on grid quantization

Abstract

The invention relates to a grid quantization-based self-adaptive unrouted path planning system and a grid quantization-based self-adaptive unrouted path planning method. The grid quantization-based self-adaptive unrouted path planning system comprises a data layer, a logic operation layer and a control layer, the logical operation layer comprises the following steps: grid division: performing uniform square grid division on a to-be-planned region; grid quantitative scoring: constructing a navigation body trafficability scoring rule base, performing quantitative scoring in each grid, and finally superposing scores of various data; grid tile generation: generating tiles from the data after grid quantization to form multiple pieces of multi-level multi-resolution tile data; grid path planning: performing path calculation by utilizing a shortest path planning algorithm based on the tile grid data, and dynamically loading required tile data during path finding; and the application layer is used for carrying out operation control on the logic operation layer and carrying out result output control on an upper-layer application. The problems of single grid granularity, insufficient environment feature expression, weak dynamic planning capability and the like of a path planning technology are solved.

Description

Technical Field

[0001] This invention relates to an intelligent navigation and path planning technology, and more particularly to a general path planning system based on a grid quantization method, which is suitable for dynamic path generation under terrain conditions without road networks. Background Technology

[0002] I. Current Status of Related Technology Development For scenarios without clearly defined roads, such as off-road and field operations, researchers have developed path planning methods based on grid maps and Voronoi diagrams. These techniques use environmental modeling and obstacle detection to search for feasible paths in roadless areas. Key features include: 1) employing environmental discretization; 2) strong environmental adaptability; and 3) high path safety.

[0003] II. Problems with existing related technologies 2.1 Limitations of Mesh Quantization Methods: While existing mesh quantization methods have solved environmental modeling problems to some extent, they still have significant shortcomings: Uniform mesh granularity: 1) Fixed-size meshes are difficult to adapt to planning requirements of different precision; 2) In complex environments, either the computational load is too large or the precision is insufficient; 3) Data reconstruction is required for each navigation body separately.

[0004] Insufficient representation of environmental features: 1) Traditional grids mainly represent obstacle information; 2) Lack of quantitative representation of multi-dimensional features such as terrain slope and surface adhesion coefficient.

[0005] 2.2 Weak dynamic programming capabilities: Existing related technologies show significant shortcomings in coping with dynamic environmental changes. Poor real-time performance: 1) Slow replanning response speed; 2) Unable to meet the real-time requirements of high-speed mobile scenarios. Summary of the Invention

[0006] This application proposes an adaptive no-path planning system and method based on grid quantization to quantize roadless areas, comprehensively consider the maneuverability of the navigation vehicle to score the quantized grid, and perform A* path calculation on the quantized and scored grid map. This aims to solve the problems of single grid granularity, insufficient expression of environmental features, and weak dynamic planning ability in existing path planning technologies. At the same time, a grid dynamic update mechanism is added to solve the problem of weak dynamic planning ability.

[0007] The technical solution of this invention is as follows: An adaptive no-path planning system based on grid quantization, the system comprising: Data Layer: The data layer provides a data reading interface for the logic operation layer; it includes: road network data, digital elevation data, geological data, topographic data, dynamic data, and other data; Logical operation layer: The logical operation layer includes mesh generation, mesh quantization and scoring, mesh tile generation, and path planning; details are as follows: Grid division: Divide the area to be planned into uniform square grids; Grid-based quantitative scoring: Construct a navigation vehicle passability scoring rule library, including scoring rules for navigation vehicles under each type of data, which can be dynamically adjusted; For each type of data, quantitative scoring is performed in each grid, and the scoring is based on the navigation vehicle passability scoring rule library. The scores are normalized, and the scores of multiple data are finally superimposed and then normalized again. Mesh tile generation: Introducing a tile data organization method, the complete mesh data is divided into sub-data regions of the same size, and the unpathed mesh region is divided into four tile data sub-regions. Each tile data is a square region containing the same number of meshes; the quantized mesh data is used to generate tiles, forming multiple tile data at multiple levels and resolutions. Grid path planning: Based on tile grid data, the shortest path planning algorithm is used to calculate the path. The required tile data is dynamically loaded during path finding, and the tile data is released from memory after it is used. Application layer: The application layer controls the operations of the logic operation layer and controls the output of results from the upper-layer applications.

[0008] Furthermore, the data layer specifically includes: Road network data: The road network data used for route planning includes attributes such as road grade, road width, and road ancillary information; Digital elevation data: used to calculate the elevation information of each grid, and further to derive the slope and aspect information of each grid; Geological data: In the field, navigation devices are highly sensitive to the geological environment; Topographic data: including buildings, rivers, lakes, shrubs, fences, and railings, which are obstacles that prevent navigation from passing through; Dynamic data: Dynamic data refers to data that may affect the passage of navigation vehicles due to accidental events, including: bridge collapses and road collapses caused by natural disasters; Other data: Other factors that may affect the passage of navigation vehicles.

[0009] Furthermore, the logic operation layer specifically includes: Grid division: The side length of each grid cell is 1-1.5 times the length of the navigation body; Grid-based quantitative scoring: Quantitative scoring, with scores normalized to [0,1]. Higher scores indicate better passability. The navigation body passability scoring rule base is formed based on experience or technical specialist research and scoring. Mesh tile generation: Enables dynamic loading of tile data required for path planning calculations, while also allowing timely release of unnecessary tile data; Grid path planning: Using the A* algorithm as the shortest path planning algorithm.

[0010] Furthermore, the meshing adopts an adaptive meshing rule, using a quadtree partitioning method. It recursively divides the two-dimensional space into four quadrants for organization, which facilitates dynamic meshing or aggregation based on different navigation body shape parameters. The same set of meshes can be used to adapt to multiple navigation bodies through meshing and aggregation. In practical applications, a relatively fine meshing can be performed on the entire data. For different navigation bodies or when navigation body parameters change, only adaptive adjustments need to be made based on the data, without re-meshing. This enables a rapid response to the planning requirements of the application layer.

[0011] Furthermore, the grid quantization scoring adopts a multi-data-factor fast grid quantization rule. When quantifying and scoring each grid, the following grid scoring rules are formulated: For each factor, evaluate the passability of all grids, find all grids that meet the "one-vote veto" rule, and record the passability as 0; otherwise, record the passability as 1. The evaluation results for each factor are superimposed on each grid. When the passability of any factor is 0, the passability of that grid is 0; otherwise, it is 1. All grids with a passability of 1 after stacking are re-evaluated.

[0012] Furthermore, in the generation of grid tiles, the tile data structure is as follows: Each tile contains the following data: Tile range: The latitude and longitude range covered by the tile, expressed by the latitude and longitude of the lower left and upper right points of the tile, i.e. minimum longitude, minimum latitude, maximum longitude, and maximum latitude; The index or data information of the grid contained in the tile: indicates which grids the tile covers, which can be expressed by grid index or specific grid information; Other tiles adjacent to this tile and their connectivity: Store the tile numbers around this tile and the connectivity between this tile and them. Store this information to quickly find the next tile to be loaded during pathfinding and quickly determine whether loading is necessary. The tiles need to be coded: that is, each tile has a unique number; Tile division rules: Each tile is a square area of ​​the same size and with a consistent coverage area, and its longitude and latitude are both integer multiples of the grid side length; The tile size is matched and configured according to actual operation, and the tile side length is 10-50 times the grid side length.

[0013] Furthermore, grid path planning also includes: In practical applications, these dynamic data need to be integrated into the path planning data in real time. Using the shortest path planning algorithm, the dynamic data only needs to be grid-quantized, and then spatially superimposed with the original grid-quantized results for calculation.

[0014] An adaptive no-path planning method based on grid quantization, used to implement the adaptive no-path planning system based on grid quantization as described in any one of claims 1-7, comprising: Step 1) Divide the data range without paths into grids; Step 2) Grid score each type of data within the range. The final result of each grid is a normalized passability score. The higher the score, the better the passability. Step 3) Generate tiles by merging the mesh; Step 4) Perform path planning on the scored grid to form a path from the starting point to the destination; Step 4.1) Input route planning data, including the latitude and longitude of the origin and destination, the type of navigation body, and its specific length, width, height, and weight parameters; Step 4.2) Load the tile data of the origin and destination, and use the shortest path algorithm to find the path; Step 4.3) Determine whether to load new tile data when the pathfinding traverses a new data location, and release the used tile data in a timely manner; Step 4.4) Search for the shortest path to the destination.

[0015] The beneficial effects of this invention are as follows: The key technical points of this application are the entire no-path planning technical framework, which applies adaptive grid partitioning rules to no-path planning, and proposes a fast grid quantization strategy for multiple data factors and a dynamic data update strategy for no-path planning.

[0016] This application proposes a no-path planning algorithm based on grid quantization scoring. Through steps such as grid partitioning, grid quantization, grid aggregation, and path planning, it can achieve adaptive path calculation for different navigation bodies under various data superposition environments, solving the problems of single grid granularity and insufficient expression of environmental features in traditional no-path planning. At the same time, it introduces a fast grid quantization strategy with multiple data factors, which can simplify the grid quantization calculation process and improve the path calculation efficiency. For the dynamic data update problem in practical applications, a fast dynamic data update strategy is proposed to solve the problem of weak dynamic planning capability of traditional no-path planning. Attached Figure Description

[0017] Figure 1 This is a system architecture diagram of the adaptive no-path planning method of the present invention; Figure 2 This is a flowchart of the adaptive no-path planning method of the present invention; Figure 3 This is a diagram showing the mesh division and aggregation of the present invention; Figure 4 This is a schematic diagram of the tile data organization of the present invention. Detailed Implementation

[0018] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. These embodiments are based on the technical solution of the present invention and provide detailed implementation methods and specific operating procedures. However, the scope of protection of the present invention is not limited to the following embodiments.

[0019] An adaptive no-path planning system and method based on grid quantization, such as Figure 1 As shown, the system includes: (1) Data layer The data layer includes all the data required in this application, including: road network data, digital elevation data, geological data, topographic data, dynamic data, and other data.

[0020] Road network data: The road network data used in traditional route planning must include attributes such as road grade, road width, and road ancillary information.

[0021] Digital elevation data: used to calculate the elevation information of each grid, and further to obtain the slope and aspect information of each grid. The slope can seriously affect the passability of the navigation object.

[0022] Geological data: In the field, navigation devices such as large off-road vehicles are very sensitive to the geological environment. For example, gravel and clay have different effects on the passage of navigation devices.

[0023] Topographic data: Buildings, rivers, lakes, shrubs, fences, guardrails, etc. are obstacles to the navigation object and cannot be passed.

[0024] Dynamic data: Dynamic data refers to data that may affect the passage of navigation vehicles due to occasional events, such as bridge collapses and road collapses caused by natural disasters.

[0025] Other data: Other factors that may affect the passage of navigation vehicles.

[0026] The data layer provides a data reading interface for the logic operation layer.

[0027] (2) Logical operation layer The logic operation layer mainly consists of four parts: grid partitioning, grid quantization and scoring, grid tile generation, and path planning.

[0028] 1) Grid division: The area to be planned is divided into uniform square grids. The size of the grid division determines the number of grids, which in turn determines the computational load of subsequent path planning and also affects the accuracy of path planning. The smaller the grid, the higher the accuracy. At the same time, the larger the number of grids, the more accurate the path planning. In order to ensure a balance between accuracy and computational cost, this invention proposes that the side length of the grid be 1-1.5 times the length of the navigation body.

[0029] 2) Grid-based Quantitative Scoring: A navigation vehicle passability scoring rule library is constructed. For each type of data, a quantitative score is performed within each grid, with scores normalized to [0,1]. Higher scores indicate better passability. The scoring is based on the navigation vehicle passability scoring rule library, which includes scoring rules for each type of data. For example, for off-road vehicles, dense jungle results in a score close to 0. This rule library is formed based on experience or technical specialist research and scoring, and can be dynamically adjusted to adapt to various navigation vehicles. The scores from multiple data sets are finally superimposed and then normalized.

[0030] 3) Grid Tile Generation: Drawing inspiration from the loading of electronic maps with tiles, the data after grid quantization is used to generate tiles, forming multiple tile data at multiple levels and resolutions. This allows for the dynamic loading of tile data required for path planning calculations, while also enabling the timely release of unnecessary tile data, without having to load a large amount of data at the beginning. This effectively solves the problem of insufficient memory when performing path planning calculations on a large scale.

[0031] 4) Grid path planning: Based on tile grid data, the shortest path planning algorithm such as A* is used to calculate the path. The required tile data is dynamically loaded during path finding, and the tile data is released from memory after it is used.

[0032] (3) Application layer The application layer mainly controls the operations of the logic operation layer and the output of results from the upper-layer applications.

[0033] II. Algorithm Flow The algorithm flow of this application is as follows, such as Figure 2 As shown: (1) Divide the data range without a path into a grid; (2) Grid scoring is performed on each type of data within the range. The final result of each grid is a normalized passability score. The higher the score, the better the passability. (3) It can generate tiles by merging and organizing grids; (4) Path planning is performed on the grid after scoring to form a path from the starting point to the end point.

[0034] 4.1) Input route planning data, including the latitude and longitude of the origin and destination, the type of navigation body, and its specific length, width, height, and weight parameters; 4.2) Load the tile data of the origin and destination, and use the A* algorithm for pathfinding; 4.3) When a new data location is encountered during the pathfinding traversal, determine whether new tile data needs to be loaded and release the used tile data in a timely manner; 4.4) Search for the shortest path to the destination.

[0035] III. Adaptive Mesh Generation Rules Grid partitioning is the foundation for subsequent grid quantization scoring and path calculation. To ensure that the partitioned grid can adapt to or support more navigation objects, this application adopts a quadtree partitioning method. A quadtree is a tree-like data structure whose core characteristic is that it organizes a two-dimensional space by recursively dividing it into four quadrants. This facilitates dynamic grid partitioning or aggregation based on different navigation object shape parameters. For example, using a 1-meter grid side length can accommodate a grid suitable for human walking. If the grid is used for a vehicle, assuming the vehicle's length is 4 meters, the original grid can be aggregated twice (each adjacent 4×4 grids can be aggregated into one), without needing to re-partition. For smaller navigation objects, such as a robot dog, the 1-meter grid can be further divided into 4. Figure 3 As shown.

[0036] The advantage of this approach is that the same set of grids can be used to adapt to multiple navigation bodies through grid division and aggregation. In practical applications, the entire data can be divided into a relatively fine grid. For different navigation bodies or when the parameters of the navigation body change, only adaptive adjustments need to be made based on the data, without having to redo the grid division. This allows for a rapid response to the planning requirements of the application layer.

[0037] IV. Rapid Grid Quantization Strategy Based on Multiple Data Factors To reduce unnecessary grid quantization, this invention introduces a fast grid quantization rule for multiple data factors, which can accelerate the scoring of multiple data quantizations.

[0038] When quantifying the score for each grid, a "one-vote veto" system is considered. For example, if a grid is located inside a river, its passability is 0 without considering any other factors, and no other factors need to be quantified for that grid.

[0039] Based on the above principles, the following grid scoring rules are formulated: 1) Evaluate the passability of all grids for each factor, find all grids that meet the "one-vote veto system", and record the passability as 0; otherwise, record the passability as 1. 2) The evaluation results of each factor are superimposed on each grid. When the passability of any factor is 0, the passability of that grid is 0; otherwise, it is 1. 3) Re-evaluate all grids with a passability of 1 after overlay.

[0040] By using the above rules, the workload of grid quantization scoring can be greatly reduced, thereby significantly accelerating the calculation efficiency. In actual tests, it was found that using these rules can improve the calculation efficiency by an average of about 23%.

[0041] V. Tile-based Grid Data Organization Strategy A very serious problem encountered when performing large-scale grid path calculation is that the number of grids is extremely large. When using A* or Dijkstra's algorithm to traverse the grid, memory often runs out due to loading a large number of grids at once.

[0042] This application introduces a tile data organization method, dividing the complete grid data into sub-data regions of equal size, such as... Figure 4 As shown, the unconnected grid area is divided into four tile data sub-regions, each of which is a square area containing the same number of grids.

[0043] (1) Tile data structure Each tile contains the following data: 1) Tile range: The latitude and longitude range covered by the tile, usually expressed by the latitude and longitude of the lower left and upper right points of the tile, i.e. minimum longitude, minimum latitude, maximum longitude, and maximum latitude.

[0044] 2) Index or data information of the grid contained in the tile: Indicates which grids the tile covers, which can be expressed by grid index or specific grid information (including grid code and corresponding score value, etc.).

[0045] 3) Other tiles adjacent to this tile and their connectivity: Store the tile numbers around this tile and the connectivity between this tile and them. Store this information so that the next tile to be loaded can be found quickly during pathfinding and it can be quickly determined whether loading is necessary. For example, if two adjacent tiles A and B are not connected, there is no need to search the area of ​​tile B when expanding pathfinding on tile A, and there is no need to load the data of tile B.

[0046] 4) Tiles need to be coded: that is, each tile has a unique number.

[0047] (2) Rules for dividing tiles 1) Each tile is a square area of ​​the same size and with a consistent coverage area, and its longitude and latitude are both integer multiples of the grid side length.

[0048] 2) The tiles should not be too small or too large. The ideal tile side length is 10-50 times the grid side length. VI. Dynamic Data Update Strategy for Pathless Path Planning Considering that in practical applications, terrain and landform conditions may change over time, such as beaches being affected by tides; and that sudden events may affect the navigation of the vehicle, such as landslides, wildfires and other sudden natural disasters.

[0049] These dynamic data need to be integrated into the path planning data in real time. Using the aforementioned path planning algorithm, we only need to perform grid quantization on the dynamic data and then spatially overlay it with the original grid quantization result. There is no need to re-grid quantize and score all the data, which does not have a significant impact on the complexity of the entire path planning calculation.

[0050] The above-described embodiments are merely one implementation of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention should be determined by the appended claims.

Claims

1. An adaptive no-path planning system based on grid quantization, characterized in that, The system includes: Data Layer: The data layer provides a data reading interface for the logic operation layer; it includes: road network data, digital elevation data, geological data, topographic data, dynamic data, and other data; Logical operation layer: The logical operation layer includes mesh generation, mesh quantization and scoring, mesh tile generation, and path planning; details are as follows: Grid division: Divide the area to be planned into uniform square grids; Grid-based quantitative scoring: Construct a navigation vehicle passability scoring rule library, including scoring rules for navigation vehicles under each type of data, which can be dynamically adjusted; For each type of data, quantitative scoring is performed in each grid, and the scoring is based on the navigation vehicle passability scoring rule library. The scores are normalized, and the scores of multiple data are finally superimposed and then normalized again. Mesh tile generation: Introducing a tile data organization method, the complete mesh data is divided into sub-data regions of the same size, and the unpathed mesh region is divided into four tile data sub-regions. Each tile data is a square region containing the same number of meshes; the quantized mesh data is used to generate tiles, forming multiple tile data at multiple levels and resolutions. Grid path planning: Based on tile grid data, the shortest path planning algorithm is used to calculate the path. The required tile data is dynamically loaded during path finding, and the tile data is released from memory after it is used. Application layer: The application layer controls the operations of the logic operation layer and controls the output of results from the upper-layer applications.

2. The adaptive no-path planning system based on grid quantization according to claim 1, characterized in that, The data layer specifically includes: Road network data: The road network data used for route planning includes attributes such as road grade, road width, and road ancillary information; Digital elevation data: used to calculate the elevation information of each grid, and further to derive the slope and aspect information of each grid; Geological data: In the field, navigation devices are highly sensitive to the geological environment; Topographic data: including buildings, rivers, lakes, shrubs, fences, and railings, which are obstacles that prevent navigation from passing through; Dynamic data: Dynamic data refers to data that may affect the passage of navigation vehicles due to accidental events, including: bridge collapses and road collapses caused by natural disasters; Other data: Other factors that may affect the passage of navigation vehicles.

3. The adaptive no-path planning system based on grid quantization according to claim 1, characterized in that, The logic operation layer specifically includes: Grid division: The side length of each grid cell is 1-1.5 times the length of the navigation body; Grid-based quantitative scoring: Quantitative scoring, with scores normalized to [0,1]. Higher scores indicate better passability. The navigation body passability scoring rule base is formed based on experience or technical specialist research and scoring. Mesh tile generation: Enables dynamic loading of tile data required for path planning calculations, while also allowing timely release of unnecessary tile data; Grid path planning: Using the A* algorithm as the shortest path planning algorithm.

4. The adaptive no-path planning system based on grid quantization according to claim 1, characterized in that, The meshing adopts an adaptive meshing rule and uses a quadtree partitioning method. It recursively divides the two-dimensional space into four quadrants for organization, which facilitates dynamic meshing or aggregation based on different navigation body shape parameters. The same set of meshes can be used to adapt to multiple navigation bodies through meshing and aggregation. In practical applications, a relatively fine meshing can be performed on the entire data. For different navigation bodies or when the navigation body parameters change, only adaptive adjustments need to be made based on the data, without re-meshing. This can quickly respond to the planning requirements of the application layer.

5. The adaptive no-path planning system based on grid quantization according to claim 1, characterized in that, The grid quantization scoring adopts a multi-data-factor fast grid quantization rule. When quantifying and scoring each grid, the following grid scoring rules are defined: For each factor, evaluate the passability of all grids, find all grids that meet the "one-vote veto" rule, and record the passability as 0; otherwise, record the passability as 1. The evaluation results for each factor are superimposed on each grid. When the passability of any factor is 0, the passability of that grid is 0; otherwise, it is 1. All grids with a passability of 1 after stacking are re-evaluated.

6. The adaptive no-path planning system based on grid quantization according to claim 1, characterized in that, During grid tile generation, the tile data structure is as follows: Each tile contains the following data: Tile range: The latitude and longitude range covered by the tile, expressed by the latitude and longitude of the lower left and upper right points of the tile, i.e. minimum longitude, minimum latitude, maximum longitude, and maximum latitude; The index or data information of the grid contained in the tile: indicates which grids the tile covers, which can be expressed by grid index or specific grid information; Other tiles adjacent to this tile and their connectivity: Store the tile numbers around this tile and the connectivity between this tile and them. Store this information to quickly find the next tile to be loaded during pathfinding and quickly determine whether loading is necessary. The tiles need to be coded: that is, each tile has a unique number; Tile division rules: Each tile is a square area of ​​the same size and with a consistent coverage area, and its longitude and latitude are both integer multiples of the grid side length; The tile size is matched and configured according to actual operation, and the tile side length is 10-50 times the grid side length.

7. The adaptive no-path planning system based on grid quantization according to claim 1, characterized in that, Grid path planning also includes: In practical applications, these dynamic data need to be integrated into the path planning data in real time. Using the shortest path planning algorithm, the dynamic data only needs to be grid-quantized, and then spatially superimposed with the original grid-quantized results for calculation.

8. An adaptive no-path planning method based on grid quantization, characterized in that, For implementing the adaptive no-path planning system based on grid quantization as described in any one of claims 1-7, comprising: Step 1) Divide the data range without paths into grids; Step 2) Grid score each type of data within the range. The final result of each grid is a normalized passability score. The higher the score, the better the passability. Step 3) Generate tiles by merging the mesh; Step 4) Perform path planning on the scored grid to form a path from the starting point to the destination; Step 4.1) Input route planning data, including the latitude and longitude of the origin and destination, the type of navigation body, and its specific length, width, height, and weight parameters; Step 4.2) Load the tile data of the origin and destination, and use the shortest path algorithm to find the path; Step 4.3) Determine whether to load new tile data when the pathfinding traverses a new data location, and release the used tile data in a timely manner; Step 4.4) Search for the shortest path to the destination.

Images