Two-stage self-adaptive complex terrain global path planning method

Abstract

The invention discloses a two-stage self-adaptive complex terrain global path planning method, which comprises the following steps of: calculating movement cost by taking a grid unit corresponding to the finest resolution as a unit and fusing gradient, ground cover, roughness and topographic relief; determining the passing cost of the region; a path planned in the first stage is used as a reference path of the second stage, the problems of efficiency and precision of the second path planning stage are comprehensively considered, an adaptive resolution mechanism is adopted, and adaptive path end point judgment is adopted for each middle section of path in the second stage, so that the flexibility and expandability of the path are enhanced, and the path planning efficiency is improved. And accurate and rapid path planning of a complex terrain path is realized.

Description

Technical Field

[0001] The present invention relates to the field of path planning, and particularly to a two-stage adaptive global path planning method for complex terrains. Background Art

[0002] With the continuous expansion of the application fields of mobile robots, their working environments are no longer limited to simple indoor places and are gradually expanding to complex outdoor terrains. Working in complex terrain environments, mobile robots can replace humans to complete dangerous or arduous tasks, which has significant benefits. In complex environments such as mountains, deserts, or disaster areas, robots can perform exploration, rescue, or material transportation tasks, avoiding humans being exposed to risks such as cliffs and landslides and ensuring safety. At the same time, they can continuously operate without being restricted by fatigue, replacing humans to complete long-term inspections or data collection and improving efficiency. In addition, robots can also operate in harsh climates or toxic environments, such as volcanic monitoring or chemical leak detection, replacing humans to avoid health risks, thereby increasing the success rate of tasks and reducing labor costs. Path planning in complex terrain environments refers to finding a feasible and optimal path from a starting point to an ending point under the condition of no road network. It is particularly difficult to perform path planning in complex terrain environments. Traditional path planning methods mostly focus on path design in structured environments, while in complex natural environments, these methods often cannot effectively cope with variable ground features. Therefore, the problem of path planning for complex terrain environments has become a difficult problem that researchers urgently need to solve. Currently, path planning algorithms face the following problems in complex terrains: (1) Low computational efficiency: Single-resolution search leads to a large number of nodes, especially in large-scale complex terrains, which takes a significant amount of time; (2) Poor terrain adaptability, resulting in low path feasibility; (3) Single cost evaluation, where traditional evaluation functions only consider distance and ignore the impact of terrain on movement costs; (4) Lack of dynamic adjustment: Existing two-stage path planning algorithms lack a multi-stage resolution adaptive mechanism and are difficult to balance accuracy and efficiency. Summary of the Invention

[0003] Object of the Invention: The present invention aims to provide a two-stage adaptive global path planning method for complex terrains that integrates terrain parameters.

[0004] Technical Solution: The two-stage adaptive global path planning method described in the present invention includes the following steps: (1) Match the elevation data map and the surface cover type data map of the path planning area, define the resolution of the matched map as the finest resolution, and calculate the slope, roughness, and terrain undulation of the grid cells at the finest resolution; (2)Based on the slope, roughness, and terrain undulation degree of the grid cells obtained in step (1), determine the slope cost, roughness cost, and terrain undulation degree cost of the grid cells corresponding to the finest resolution, and set the surface cover cost according to the regional type; (3)In the first stage, use the bidirectional A-star algorithm for fast path planning; determine the first-stage retrieval resolution according to the path planning area and the preset fast retrieval space, calculate the traversal cost of the grid area corresponding to the first-stage retrieval resolution based on the slope cost, roughness cost, terrain undulation degree cost of the grid cells corresponding to the finest resolution obtained in step (2), and the set surface cover cost, and obtain the optimal path set of the first stage using the bidirectional A-star algorithm; (4)In the second stage, use the adaptive A-star algorithm for path planning; construct an adaptive path resolution for the optimal path obtained in step (3), calculate the traversal cost of the grid area corresponding to the adaptive path resolution, use the adaptive A-star algorithm, introduce a path adaptive buffer attenuation function, and perform adaptive path end point judgment on each intermediate path segment to complete the path planning.

[0005] Further, in step (1), convert the data of the elevation data map and the surface cover type data map to the same coordinate system. If the resolution of the elevation data map is lower than that of the surface cover type data map, resample the elevation data map using bilinear interpolation; otherwise, resample the surface cover type data map using the nearest neighbor method.

[0006] Further, in step (1), with the corresponding grid cell to be solved as the center, expand to obtain a 3×3 nine-grid area, and the elevation value matrix corresponding to the nine-grid is ; The slope of the grid cell corresponding to the finest resolution is ; Among them, represents the slope in the row direction, represents the slope in the column direction, is the finest resolution in the row direction after map matching, is the finest resolution in the column direction after map matching, and S represents the slope value of the grid cell; The roughness R of the grid cell corresponding to the finest resolution is ; Among them, is the elevation value represented by the i th grid, is the average elevation value of the nine-grid area; The terrain undulation degree U of the grid cell corresponding to the finest resolution is ; Among them, is the maximum elevation value within the nine - grid area, is the minimum elevation value within the nine - grid area.

[0007] Furthermore, in step (2), the slope cost of the grid cell corresponding to the finest resolution is ; Among them, is the slope of the current driving grid, is the maximum allowable passing slope of the mobile robot; The roughness cost is ; Among them, R is the roughness of the nine - grid area; The terrain undulation cost is ; Among them, and respectively represent the maximum elevation value and the minimum elevation value within the nine - grid area, is the maximum elevation of the elevation within the map area, is the minimum elevation of the elevation within the map area.

[0008] Furthermore, in step (2), the surface cover cost is ; Among them, if the surface cover type of the grid cell corresponding to the finest resolution is an obstacle area, or a water area, or a tree area, it is classified as an impassable area; if the surface cover type of the grid cell corresponding to the finest resolution is a crop or building area, it is classified as a difficult - to - pass area; if the surface cover type of the grid cell corresponding to the finest resolution is bare ground or grassland, it is classified as a passable area.

[0009] Furthermore, in step (3), the resolution of the first - stage retrieval is ; ; Among them, Re c- x represents the resolution in the row direction of the first - stage retrieval, Re c- y represents the resolution in the column direction of the first - stage retrieval, mIndicating the search space when the search space is reduced to the finest resolution , indicating rounding up; The passing cost of the grid area corresponding to the retrieval resolution in the first stage is ; Among them, , , , are weight coefficients, and ; is the slope cost of the area, is the surface cover cost of the area, is the roughness cost of the area, is the terrain undulation cost of the area.

[0010] Furthermore, the slope cost of the area is ; ; Among them, , are weight coefficients, ; N is the number of grid cells corresponding to the finest resolution contained in the grid area corresponding to the retrieval resolution in the first stage, is the average value of the slope costs of all grid cells corresponding to the finest resolution in this area, is the slope cost of the th grid cell corresponding to the finest resolution; The surface cover cost of the area is ; ; Among them, , are weight coefficients, ; is the average value of the surface cover costs of all grid cells corresponding to the finest resolution in this area, is the surface cover cost of the th grid cell corresponding to the finest resolution; The roughness cost of the area is ; ; Among them, , are weight coefficients, ; is the average roughness cost of the grid cells corresponding to all the finest resolutions in this area, is the th roughness cost of the grid cell corresponding to the finest resolution; The terrain undulation cost of the area is ; ; Among them, , are weight coefficients, ; is the average terrain undulation cost of the grid cells corresponding to all the finest resolutions in this area, is the th terrain undulation cost of the grid cell corresponding to the finest resolution.

[0011] Furthermore, in step (4), an adaptive path resolution is constructed for the optimal path obtained in step (3) as ; ; Among them, is the adaptive resolution in the row direction, is the adaptive resolution in the column direction, is the maximum passing cost at the first-stage retrieval resolution; is the passing cost of the path at the first-stage retrieval resolution, represents the maximum number of grid cells corresponding to the finest resolution in the grid area corresponding to a second-stage adaptive resolution in the row direction or column direction; represents rounding down.

[0012] Furthermore, in step (4), the path adaptive buffer attenuation function f(d) is ; Among them, c is a custom coefficient, D is the distance from the end point of the last path segment to the start point of the first path segment , d is the distance from the end point of the current path segment to the start point of the first path segment .

[0013] Beneficial effects: Compared with the prior art, the present invention has the following remarkable advantages: 1. The present invention calculates the movement cost by taking the grid cell corresponding to the finest resolution as the unit and integrating the slope, surface cover, roughness, and terrain undulation; 2. The grid areas corresponding to the resolutions in the first stage and the second stage of the present invention are both composed of the grid cells corresponding to the finest resolution. The overall traversability and local volatility of the terrain are comprehensively characterized by the statistical aggregation of the cost mean and standard deviation of the grid cells within the area, so as to calculate the traversal cost of the area; 3. The present invention takes the path planned in the first stage as the reference path for the second stage, comprehensively considers the problems of efficiency and accuracy in the second path planning stage, adopts an adaptive resolution mechanism, and uses an adaptive path end point judgment for each intermediate path segment in the second stage, enhancing the flexibility and scalability of the path, and achieving accurate and fast path planning for complex terrain paths. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 is a flowchart of the present invention; Figure 2 is a schematic structural diagram of a nine-square grid. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0015] The present invention will be further described below with reference to the accompanying drawings.

[0016] The two-stage adaptive global path planning method for complex terrain according to the present invention includes the following steps: (1) Match the elevation data map and the surface cover type data map of the path planning area, define the resolution of the matched map as the finest resolution, and calculate the slope, roughness, and terrain undulation of the grid cells at the finest resolution.

[0017] Convert the data of the elevation data map and the surface cover type data map into the same coordinate system, resample the low-resolution map to match the high-resolution map. Among them, bilinear interpolation is used for the elevation data, and the nearest neighbor method is used for the surface cover type data. Then, crop the data according to the boundary range of the research area to ensure that the spatial ranges of the two maps are the same and completely covered. The row represents the east-west direction, and the column represents the north-south direction. The resolution of the processed map is the finest resolution, and the finest resolution in the row direction is and the finest resolution in the column direction is .

[0018] According to the coordinates of the path start point (srow, scol) and the end point (grow, gcol), determine the minimum row coordinate (minrow), minimum column coordinate (mincol), maximum row coordinate (maxrow), and maximum column coordinate (maxcol) of the target area, and determine the range to be rasterized. Among them: ; In the above formula, K is a user-defined expansion coefficient, is to take the absolute value. For example, is the absolute value of.

[0019] Generate the minimum row coordinates with the row resolution as the step size (minrow) to the maximum row coordinates (maxrow) of the discrete sequence; generate the minimum column with the column resolution as the step size (mincol) to the maximum column (maxcol) of the discrete sequence.

[0020] When the mobile robot is moving, it is mainly affected by several nearby areas centered on its own position. Therefore, the slope, roughness, and terrain undulation are calculated by expanding around the center of a grid area with the finest resolution to form a 3×3 nine-grid area. The nine-grid area is shown in Figure (2), where the letters in the nine-grid represent elevation values, and the elevation value matrix corresponding to the nine-grid is .

[0021] The slope of the grid cell corresponding to the finest resolution is ; where, represents the slope in the row direction, represents the slope in the column direction, is the finest resolution in the row direction after map matching, is the finest resolution in the column direction after map matching, and S represents the slope value of the grid cell; The roughness R of the grid cell corresponding to the finest resolution is ; where, is the i elevation value represented by the th grid,

[0022] Terrain undulation refers to the difference between the maximum elevation and the minimum elevation of an area. The terrain undulation of the grid cell corresponding to the finest resolution U is ; where, is the maximum elevation value within the nine-grid area, is the minimum elevation value within the nine-grid area.

[0023] (2)Determine the slope cost, roughness cost, and terrain undulation cost of the grid cells corresponding to the finest resolution based on the slope, roughness, and terrain undulation of the grid cells obtained in step (1), and set the surface cover cost according to the region type.

[0024] When a mobile robot performs path planning in complex terrain, its driving cost is mainly affected by slope, surface cover, roughness, and terrain undulation. For each influencing factor, its cost is designed according to its traversability, and its value range is [0, 1]. The smaller the cost, the greater the traversability.

[0025] The moving difficulty of the mobile robot increases rapidly with the increase of the slope. When the slope is greater than the moving slope threshold of the robot, the robot cannot pass through the grid cell area. The slope cost of the grid cells corresponding to the finest resolution cost S is ; where is the slope of the current driving grid, is the maximum allowable passing slope of the mobile robot.

[0026] The roughness cost is ; where R is the roughness of the nine-grid area.

[0027] The terrain undulation cost is ; where and represent the maximum elevation and minimum elevation in the nine-grid area respectively, is the maximum elevation in the map area, is the minimum elevation in the map area.

[0028] Divide an area into seven regions according to the surface cover type: obstacle region, water area, trees, crops, building area, bare ground, grassland. Consider the obstacle region, water area, and tree region as non-traversable regions; consider the crop and building areas as difficult-to-traverse regions; consider the bare ground and grassland as traversable regions. Define the surface cover cost as ; Among them, if the land cover type of the grid cell corresponding to the finest resolution is an obstacle area, a water area, or a tree area, it is classified as an impassable area; if the land cover type of the grid cell corresponding to the finest resolution is a crop or building area, it is classified as a difficult-to-pass area; if the land cover type of the grid cell corresponding to the finest resolution is bare ground or grassland, it is classified as a passable area.

[0029] (3) In the first stage, the bidirectional A* algorithm is used for rapid path planning; according to the path planning area and the preset rapid retrieval space, the retrieval resolution of the first stage is determined, and based on the slope cost, roughness cost, terrain undulation cost, and set land cover cost of the grid cell corresponding to the finest resolution obtained in step (2), the passage cost of the grid area corresponding to the retrieval resolution of the first stage is calculated, and the optimal path set of the first stage is obtained by using the bidirectional A* algorithm.

[0030] Let the resolution in the row direction of the retrieval resolution of the first stage be , and let the resolution in the column direction be .

[0031] The search space of the map at the finest resolution is ; where X is the distance in the row direction of the map, Y is the distance in the column direction of the map, represents the finest resolution in the row direction, represents the finest resolution in the column direction.

[0032] The search space of the map at the retrieval resolution of the first stage is ; where X is the distance in the row direction, Y is the distance in the column direction, represents the resolution in the row direction of the first stage retrieval, represents the resolution in the column direction of the first stage retrieval, and ; Then = , in order to achieve the purpose of rapid path planning in the first stage, it is necessary to reduce the search space. Let the when the search space needs to be reduced to the search space of the finest resolution of the map, then ; Then , , since the grid area corresponding to the retrieval resolution of the first stage is composed of the grid cells corresponding to the finest resolution, the following operation is performed: ; ; Indicates rounding up.

[0033] The passing cost of the grid area corresponding to the first-stage retrieval resolution is ; wherein, , , , are weight coefficients, and ; is the slope cost of the area, is the surface cover cost of the area, is the roughness cost of the area, is the terrain undulation cost of the area.

[0034] The slope cost of the area is ; ; wherein, , are weight coefficients, ; N is the number of grid cells corresponding to the finest resolution contained in the grid area corresponding to the first-stage retrieval resolution, is the average value of the slope costs of all grid cells corresponding to the finest resolution in this area, is the slope cost of the th grid cell corresponding to the finest resolution; The surface cover cost of the area is ; ; wherein, , are weight coefficients, ; is the average value of the surface cover costs of all grid cells corresponding to the finest resolution in this area, is the surface cover cost of the th grid cell corresponding to the finest resolution; The roughness cost of the area is ; ; wherein, , is the weight coefficient, ; is the average roughness cost of grid cells corresponding to all the finest resolutions in this area, is the th roughness cost of the grid cell corresponding to the finest resolution; The terrain undulation cost of the area is ; ; Among them, , are the weight coefficients, ; is the average terrain undulation cost of grid cells corresponding to all the finest resolutions in this area, is the th terrain undulation cost of the grid cell corresponding to the finest resolution.

[0035] The evaluation function for the first-stage path planning is designed as follows: ; ; ; Among them, the n node is the current grid node, is the estimated cost from the current grid node to the target node, is the actual cost from the starting point to the previous grid node. is the distance from the previous grid node to the current grid node; is the distance from the current grid node to the end point; is the average cost of the grid area corresponding to the unpassed first-stage retrieval resolution.

[0036] Define the starting point S and the end point G, and initialize two lists and , which are used to store the nodes to be explored starting from the starting point and the end point respectively. Initialize two lists S and G, which are used to store the nodes that have been processed starting from their respective starting points. These two lists are empty when initialized.

[0037] Forward search: End the search when OpenSet_S is empty. When OpenSet_S is not empty, select the node with the smallest value from OpenSet_S as the current node and set it as Current_S. Calculate the evaluation function of all adjacent path nodes of Current_S , if the adjacent path nodes are not in OpenSet_S, add the adjacent nodes to OpenSet_S and record the predecessor of this node as Current_S. Remove Current_S from OpenSet_S and add it to CloseSet_S.

[0038] Backward search: End the search when OpenSet_G is empty. When OpenSet_G is not empty, select the node with the minimum value in OpenSet_G as the current node and set it as Current_G. Calculate the evaluation function of all adjacent path nodes of Current_G , if the adjacent path nodes are not in OpenSet_G, add the adjacent nodes to OpenSet_G and record the predecessor of this node as Current_G. Remove Current_G from OpenSet_G and add it to CloseSet_G.

[0039] During each search process, it is necessary to check whether there are nodes that appear in both CloseSet_S and CloseSet_G lists at the same time. If they appear in both lists at the same time, it means that the forward search and the backward search have met, and the search stops at this time.

[0040] When the meeting point is found, form a complete path by backtracking the forward path and the backward path. Forward path: Start from the starting point and backtrack along the parent nodes to the meeting node. Backward path: Start from the end point and backtrack along the parent nodes to the meeting node. When the final path is found, return the merged path as the final result.

[0041] Take the optimal path planned by the first-stage retrieval resolution map that can reflect the fine-scale geomorphic situation as the reference planned path for the second stage, and create a path collection , where is the th segment of the path planned by the first-stage retrieval resolution map. is the first segment of the path planned by the first-stage retrieval resolution, is the last segment of the path planned by the first-stage retrieval resolution.

[0042] (4) In the second stage, use the adaptive A* algorithm for path planning; construct an adaptive path resolution for the optimal path obtained in step (3), calculate the traversal cost of the grid area corresponding to the adaptive path resolution, use the adaptive A* algorithm, introduce a path adaptive buffer attenuation function, and perform adaptive path end point judgment on each intermediate segment of the path to complete the path planning.

[0043] To balance the accuracy and efficiency of the second-stage path planning, it is necessary to design the resolution of the second stage according to the cost of the first-stage path planning. This paper proposes an adaptive fine-scale resolution mechanism. The adaptive path resolution design for each path is as follows: ; ; Among them, is the adaptive resolution in the row direction, is the adaptive resolution in the column direction, is the maximum passing cost at the first-stage retrieval resolution; is the passing cost of the path at the first-stage retrieval resolution, t represents the maximum number of grid cells corresponding to the finest resolution in the row direction or column direction of the grid area corresponding to the second-stage adaptive resolution; represents rounding down.

[0044] The passing cost of the grid area corresponding to the second-stage adaptive resolution is ; Among them,, , , , are weight coefficients, and ; is the slope cost of the area, is the surface cover cost of the area, is the roughness cost of the area, is the terrain undulation cost of the area.

[0045] The second stage adopts the adaptive A* algorithm evaluation function design as follows: ; ; ; Among them, the n node is the current grid node, is the estimated cost from the current grid node to the target node, is the actual cost from the starting point to the previous grid node. is the distance from the previous grid node to the current grid node; is the distance from the current grid node to the end point. is the average cost of the unpassed grid area corresponding to the second-stage adaptive resolution of this section. The path planning is carried out for the path starting from the first path in turn.

[0046] Define the starting point S and the ending point G, and initialize the list used to store the nodes to be explored starting from the starting point. Initialize the list , used to store the nodes that have been processed starting from the starting point. When initializing, the CLOSE list is empty.

[0047] Forward search: End the search when OPEN is empty. When OPEN is not empty, select from OPEN . If the adjacent path node is not in CLOSE and not in OPEN, add the adjacent path node to OPEN, record the predecessor of this node as Current, remove Current from OPEN, and add it to CLOSE.

[0048] Adaptive end point judgment: For each intermediate path segment , introduce a path adaptive buffer decay function, and use the adaptive A* algorithm for path planning. The buffer decay function is designed as follows: ; where c is a user-defined coefficient, D is the last path segment the distance from the end point to the first path segment from the starting point, d is the distance from the end point of the current path segment to the first path segment from the starting point, is the absolute value of. When the buffer decay function is satisfied, that is, the distance from the current path node to the end point of this path segment , it is considered that the end point has been reached, and this path node is regarded as the end point of this path segment.

[0049] Loop and end: Take the end point of the current path as the starting point of the next path segment, and perform adaptive resolution path planning for each path segment in sequence until the planning of the last path segment is completed.

Claims

1. A two-stage adaptive complex terrain global path planning method, characterized in that: The following steps are involved: (1) Match the elevation data map and the surface cover type data map of the path planning area, define the resolution of the matched map as the finest resolution, and calculate the slope, roughness, and terrain undulation of the grid unit at the finest resolution; (2) according to the slope, roughness and terrain relief of the grid cell obtained in step (1), determine the slope cost, roughness cost and terrain relief cost of the grid cell corresponding to the finest resolution, and set the surface cover cost according to the area type; (3) In the first stage, the bidirectional A-star algorithm is used for fast path planning. According to the path planning area and the preset fast search space, the first stage search resolution is determined. According to the slope cost, roughness cost, terrain undulation cost and set surface cover cost of the grid unit corresponding to the finest resolution obtained in step (2), the travel cost of the grid area corresponding to the first stage search resolution is calculated, and the bidirectional A-star algorithm is used to obtain the optimal path set of the first stage. (4) In the second stage, the adaptive A-star algorithm is used for path planning. The adaptive path resolution is constructed for the optimal path obtained in step (3), and the travel cost of the grid area corresponding to the adaptive path resolution is calculated. The adaptive A-star algorithm is used, and the path adaptive buffer attenuation function is introduced. The adaptive path endpoints of each intermediate path segment are determined to complete the path planning.

2. The two-stage adaptive complex terrain global path planning method according to claim 1 is characterized in that: In step (1), the data of the elevation data map and the land cover type data map are converted to the same coordinate system. If the resolution of the elevation data map is lower than that of the land cover type data map, the elevation data map is resampled using bilinear interpolation; otherwise, the land cover type data map is resampled using the nearest neighbor method.

3. The two-stage adaptive complex terrain global path planning method according to claim 1 is characterized in that: In step (1), the corresponding grid unit to be solved is taken as the center, and a 3×3 nine-square grid area is obtained. The elevation value matrix corresponding to the nine-square grid is ; The slope of the grid cell corresponding to the finest resolution is ； in, represents the slope in the row direction, represents the slope in the column direction, The finest resolution in the direction of travel after map matching, It is the finest resolution in the column direction after map matching, and S represents the slope value of the grid cell; The roughness R of the grid cell corresponding to the finest resolution is ； in, It is i The elevation value represented by the grid. is the average elevation value of the nine-square grid area; The topographic relief of the grid cell corresponding to the finest resolution U for ； in, is the maximum elevation value in the Jiugong grid area, It is the minimum elevation value in the nine-square grid area.

4. The two-stage adaptive complex terrain global path planning method according to claim 3 is characterized in that: In step (2), the slope cost of the grid cell corresponding to the finest resolution is for ； in, is the slope of the current driving grid, The maximum slope allowed for the mobile robot; Roughness cost for ； Among them, R is the roughness of the nine-square grid area; Terrain relief cost for ； in, and They represent the maximum and minimum elevation values ​​in the nine-square grid area respectively. is the maximum value of the elevation in the map area, The minimum elevation value in the map area.

5. The dual-stage adaptive complex terrain global path planning method according to claim 4 is characterized in that: In step (2), the cost of ground cover cost C for ； Among them, if the surface cover type of the grid unit corresponding to the finest resolution is an obstacle area, water area or tree area, it is divided into an inaccessible area; if the surface cover type of the grid unit corresponding to the finest resolution is a crop or building area, it is divided into a difficult-to-pass area; if the surface cover type of the grid unit corresponding to the finest resolution is bare ground or grassland, it is divided into a passable area.

6. The dual-stage adaptive complex terrain global path planning method according to claim 5 is characterized in that: In step (3), the first stage retrieval resolution is ； ； in, Indicates the resolution in the row direction of the first stage retrieval, Indicates the resolution in the column direction of the first stage retrieval, m Represents the search space when the search space is reduced to the finest resolution , Indicates rounding up; The first stage retrieves the travel cost of the grid area corresponding to the resolution for ； in, , , , is the weight coefficient, and ; is the slope cost of the region, is the land cover cost of the region, is the roughness cost of the region, is the terrain relief cost of the area.

7. The dual-stage adaptive complex terrain global path planning method according to claim 6 is characterized in that: Slope cost of the area for ； ； in, , is the weight coefficient, ; N is the number of grid cells corresponding to the finest resolution contained in the grid area corresponding to the first stage retrieval resolution, is the average slope cost of all grid cells corresponding to the finest resolution in the area. For the The slope cost of the grid cell corresponding to the finest resolution; Land cover cost of the region for ； ； in, , is the weight coefficient, ; is the average cost of the surface cover of all grid cells corresponding to the finest resolution in the region, For the The land cover cost of the grid cell corresponding to the finest resolution; Roughness cost of the region for ； ； in, , is the weight coefficient, ; is the average roughness cost of all grid cells corresponding to the finest resolution in the area, For the The roughness cost of the grid cell corresponding to the finest resolution; The terrain relief cost of the area for ； ； in, , is the weight coefficient, ; is the average terrain relief cost of all grid cells corresponding to the finest resolution in the area, For the The terrain relief cost of the grid cell corresponding to the finest resolution.

8. The dual-stage adaptive complex terrain global path planning method according to claim 7 is characterized in that: In step (4), the optimal path obtained in step (3) is used to construct an adaptive path with a resolution of ； ； in, is the adaptive resolution in the row direction, is the adaptive resolution in the column direction, The maximum pass cost at the first stage retrieval resolution; is the cost of the path at the resolution retrieved in the first stage, Indicates the maximum number of grid cells corresponding to the finest resolution in the row direction or column direction of a grid area corresponding to a second-stage adaptive resolution; Indicates rounding down.

9. The dual-stage adaptive complex terrain global path planning method according to claim 8, characterized in that: In step (4), the path adaptive buffer attenuation function f(d) for ； in, c is the custom coefficient, and D is the last path The end point to the first path The distance from the starting point, d From the end point of the current segment to the first segment The distance from the starting point.

Images