Lunar long-distance path planning method fusing terrain, illumination and abundance characteristics

Abstract

The invention belongs to the technical field of path planning, and discloses a moon long-distance path planning method fusing terrain, illumination and abundance characteristics, which comprises the following steps of: firstly, comprehensively analyzing a DEM image, an average illumination rate image and a rock abundance image by using a safety evaluation rule to obtain a safety map, then constructing the safety map into a tile pyramid, and finally constructing the tile pyramid by using the safety evaluation rule. The method comprises the following steps: dividing a DOM (Document Object Model) image into tiles according to layers, storing the tiles in an HDFS (Hadoop Distributed File System), performing distributed calculation on each layer of tiles from top to bottom by utilizing a Spark distributed calculation engine to obtain a global planning path, and finally performing fine periodic planning on the global planning path in combination with the DOM image.

Description

Technical Field

[0001] The present invention belongs to the technical field of path planning, and in particular relates to a lunar long-distance path planning method integrating terrain, illumination and abundance characteristics. Background Art

[0002] The moon is an important target for human deep space exploration and an ideal base and outpost for extraterrestrial space exploration. As the only natural satellite of the Earth, the moon's unique geological structure and surface characteristics record the early evolutionary history of the solar system. By sampling and analyzing the lunar surface materials, we can deeply reveal the planetary formation process and the evolution mechanism of the Earth-Moon system. In addition, the permanent shadow areas at the polar regions of the moon may be rich in volatile substances such as water ice, providing key resource guarantees for human future deep space exploration. In this context, the lunar rover, as the core equipment in the lunar exploration mission, has gradually evolved towards multifunctional integration. In the future, it will have comprehensive capabilities such as all-terrain crossing, resource exploration and utilization, manned exploration, and large-scale movement. However, the complex terrain and geomorphic environment on the lunar surface poses severe challenges to the lunar rover's exploration mission. Safe and efficient path planning has become a key prerequisite for the mobile exploration of the lunar rover.

[0003] Existing research on lunar path planning has made important progress, but there are still several shortcomings. First, most path planning studies are based on digital elevation models (DEMs) for path calculation. However, the resolution of the full-moon DEM is relatively low and cannot fully reflect the complex terrain features of the lunar surface, such as the distribution of small craters and rocks. These environmental factors that are not fully considered may pose a threat to the driving safety of the lunar rover. Secondly, as the scope of lunar exploration expands from a small local area to tens or even hundreds of kilometers, the computational complexity of path planning increases exponentially, and the computing power of a single machine is difficult to efficiently handle large-scale path planning in a complex environment. In addition, the environmental factors considered in the current path planning methods are relatively single, and most of them are optimized only based on terrain factors, and fail to comprehensively consider other important factors that may affect path safety and energy efficiency, such as light and rock abundance. Summary of the invention

[0004] The present invention proposes a lunar long-distance path planning method that integrates terrain, light, and abundance characteristics. It solves the technical problems that the environmental factors considered in current path planning methods are relatively single, and most of them are optimized only based on terrain factors, but fail to comprehensively consider other important factors such as light, rock abundance, etc. that may affect path safety and energy efficiency.

[0005] The present invention can be achieved through the following technical solutions:

[0006] A lunar long-distance path planning method that integrates terrain, illumination, and abundance features is proposed. First, a safety map is obtained by comprehensive analysis of DEM images, average illumination images, and rock abundance images using safety evaluation rules. The safety map is then constructed into a tile pyramid, which is divided into tiles and stored in HDFS. The Spark distributed computing engine is then used to perform distributed computing on each layer of tiles from top to bottom to obtain a global planning path. Finally, the global planning path is finely periodically planned in combination with DOM images.

[0007] Furthermore, when using the safety evaluation rules for comprehensive analysis, firstly, the three images of DEM, average illumination rate and rock abundance are analyzed for their accessibility, and four accessibility maps are generated accordingly. Then, the four accessibility maps are intersected to obtain a accessibility map. Finally, the accessibility map is convoluted to obtain a safety map.

[0008] The accessibility map has only two values, 0 and 1, where 0 represents that the point is an obstacle point and is not accessible; and 1 represents that the point is accessible.

[0009] Furthermore, the slope angle θ and roughness δ of each pixel in the DEM image, the average illumination rate μ of each pixel in the average illumination rate image, and the rock abundance of each pixel in the rock abundance image are calculated respectively. Four accessibility maps were generated according to the following conditions;

[0010] When the slope angle θ is greater than the first threshold, it is considered impassable and its value is 0; when it is less than the first threshold, its value is 1, so as to generate the slope passability map M slope ;

[0011] When the roughness δ is greater than the second threshold, it is considered impassable and its value is 0; when it is less than the second threshold, its value is 1, so as to generate a roughness passability map M rough ;

[0012] When the average illumination rate μ is less than the first threshold, it is considered impassable and its value is 0; when it is greater than the first threshold, its value is 1 to generate the illumination accessibility map M illumin ;

[0013] When the rock abundance φ is greater than the first threshold, it is considered impassable and its value is 0; when it is less than the first threshold, its value is 1, so as to generate the rock abundance passability map M rock .

[0014] Further, according to the following formula, the slope accessibility map M slope , Roughness passability map M rough , Light accessibility map M illumin, Rock abundance accessibility map M rock Perform intersection operation to obtain a traversability map M traverse ;

[0015] M traverse =M slope ∩M rough ∩M illumin ∩M rock

[0016] According to the following formula, the accessibility map M traverse Perform convolution calculation to obtain the security map M safety ,

[0017]

[0018] Among them, M traverse (i,j) represents the accessibility map M traverse The pixel point (i, j) on the image is K(u, v), which represents the element in the convolution kernel K. safety (i, j) represents the convolution result of the convolution kernel K at the pixel point (i, j).

[0019] Furthermore, the A* algorithm is improved from two aspects: cost function and data structure, and the improved A* algorithm is used to perform path planning on each layer of tiles.

[0020] Further, the cost function F(N) of the improved A* algorithm is calculated using the following equation:

[0021] F(N)=G(N)+(1.5-w×M safety (N))×H(N)

[0022] Among them, F(N) represents the total cost from the starting point to the end point, G(N) represents the actual cost from the starting point to the current node N, H(N) represents the estimated cost from the current node N to the end point, w represents the static weight, M safety (N) represents the accessibility of the current point N in the safety map.

[0023] Furthermore, in the improved A* algorithm, the data structure of the open table is implemented using a minimum heap and a hash table, and the data structure of the close table is implemented using a hash table.

[0024] Further, the detailed cycle planning includes the following steps:

[0025] S1. Use the Bresenham algorithm to simplify the path nodes on the global planning path to obtain a simplified path;

[0026] S2. For the simplified path, take every two adjacent path nodes as a cycle, determine the longitude and latitude range of each path segment, and extract and crop the high-resolution DOM image and security map of the corresponding area;

[0027] S3, using the CenterNet network model based on deep learning to detect small craters on the extracted high-resolution DOM images;

[0028] S4. The detected small craters are treated as obstacles and superimposed on the extracted safety map, and then the improved A* algorithm is used to re-plan the path to achieve cycle planning.

[0029] Furthermore, when the Bresenham algorithm is used to simplify the path nodes on the global planning path, starting from the starting point, each path node in the path is traversed in turn to determine whether the line between the current starting point and the traversed node passes through or is close to the obstacle point. If not, continue the traversal; if so, add the previous path node of the traversed node to the key point list, and set the previous path node as the starting point of the next round of traversal. Repeat the above process until the end point is traversed, and finally achieve path simplification.

[0030] The beneficial technical effects of the present invention are:

[0031] 1. Taking into account factors such as the slope, roughness, average illumination rate, rock abundance of the lunar terrain, and the distance and density of obstacles around the image pixels, a set of safety evaluation rules based on the lunar environment was constructed. The safety map generated based on this rule can effectively quantify the safety of various areas on the lunar surface and provide support for path planning.

[0032] 2. An improved A* algorithm is designed. This method improves the cost function and data structure on the basis of the traditional A* algorithm, reduces the time complexity and space complexity of the algorithm, and significantly improves the security and efficiency of path planning. On this basis, a distributed path planning method (DPPS-STP) based on security map tile pyramid is proposed. While ensuring path security, the calculation time is reduced, especially showing obvious advantages in large-scale path planning. Experimental results show that in long-distance path planning tasks, compared with the OC-WHT-A* algorithm in a single-machine environment, the DPPS-STP algorithm using the OC-WHT-A* of the present invention has a calculation speedup ratio of 11.53, which significantly improves the speed of path planning.

[0033] 3. Based on the Bresenham algorithm, the overall planned path is optimized to reduce the path length and turning angle. Combined with high-resolution DOM images, CenterNet is used to detect small crater obstacles, and more refined periodic path planning is performed to further improve the safety of path planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 The framework diagram of the method proposed in the present invention;

[0035] Figure 2 Schematic diagram of the structure of the convolution kernel K of the present invention;

[0036] Figure 3 Generating a flow chart for the security map of the present invention;

[0037] Figure 4 A schematic diagram of a weighted A* algorithm based on an open and closed table of a hash table data structure of the present invention;

[0038] Figure 5 A schematic diagram showing the influence of weights of different sizes on path selection according to the present invention;

[0039] Figure 6 It is a schematic diagram of the iterative calculation process of the tile pyramid model of the present invention;

[0040] Figure 7 It is a schematic diagram of a process of simplifying a path based on the Bresenham algorithm of the present invention;

[0041] Figure 8 The CenterNet crater detection schematic diagram of the present invention;

[0042] Fig. 9 Comparison of paths of different A* algorithms in a single machine environment of the present invention

[0043] Fig.10 Comparison of the path planning time of the single machine and the distributed A* algorithm in the three areas of the present invention;

[0044] Fig.11 Comparison between stand-alone and distributed path planning in three regions of the present invention; DETAILED DESCRIPTION

[0045] The specific implementation of the present invention is described in detail below with reference to the accompanying drawings and preferred embodiments.

[0046] like Figure 1 As shown, the present invention provides a lunar long-distance path planning method that integrates terrain, illumination, and abundance features. First, a safety map is obtained by comprehensive analysis of DEM images, average illumination rate images, and rock abundance images using safety evaluation rules. The safety map is then constructed into a tile pyramid, which is divided into tiles by layer and stored in HDFS. The spark distributed computing engine is then used to perform distributed computing on each layer of tiles from top to bottom to obtain a global planning path. Finally, the global planning path is finely periodically planned in combination with the DOM image.

[0047] The details are as follows:

[0048] Step 1: Build a security map

[0049] like Figure 3 As shown in the figure, the safety map is generated by comprehensively analyzing the DEM image, the average illumination rate image and the rock abundance image using the safety evaluation rules. First, considering the lunar slope, roughness, average illumination rate and rock abundance factors, the three images are analyzed for accessibility respectively to generate four accessibility maps of the corresponding factors. Then, the four accessibility maps are intersected to obtain a accessibility map. Finally, the accessibility map is convoluted to obtain the safety map. In particular, these accessibility maps have only two values, 0 and 1, where 0 represents that this point is an obstacle point and is not accessible; 1 represents that this point is accessible.

[0050] (1) Slope angle

[0051] The slope angle is an important parameter of the lunar surface topography, which directly reflects its steepness. Too large a slope will not only reduce the driving stability of the lunar rover, but may also cause the risk of sliding or rolling. This study uses the 8-neighborhood method to calculate the slope, and the calculation formula for the slope angle θ is as follows:

[0052]

[0053] Among them, f x and f y Respectively represent the gradient of the DEM image in the horizontal and vertical directions, H i It represents the elevation value of the ith grid among the surrounding 9 grids, and Cellsize is the size of the image grid.

[0054] Slope accessibility map M slope It is generated by the following conditions: when the slope angle θ is greater than the threshold, it is considered impassable and its value is 0; when it is less than the threshold, its value is 1.

[0055] (2) Roughness

[0056] Roughness is the degree to which the elevation points in the area deviate from the fitting plane, reflecting the undulation of the terrain. Excessive roughness will affect the safety and stability of the lunar rover. The calculation formula of roughness δ is as follows:

[0057]

[0058] in It represents the average value of all elevation values ​​in the above 9 grids.

[0059] The roughness traversability map M is generated by the following conditions: rough: When the roughness δ is greater than the threshold, it is marked as 0, indicating that it is impassable; otherwise, it is marked as 1.

[0060] (3) Average light rate

[0061] Lighting is a necessary condition for the lunar rover to keep warm and have sufficient solar energy. The present invention uses the annual average illumination rate data of the moon. In actual missions, the sunlight will change over time. Since the duration of long-distance patrol missions is relatively long, the average illumination rate μ can reflect the actual illumination conditions during the mission to a certain extent, which can be obtained based on the average illumination rate image analysis. By limiting the average illumination rate, the areas with poor illumination conditions on the moon can be effectively avoided.

[0062] If an area is in the shadow for a long time, its average illumination rate is less than the threshold, and the illumination accessibility map M illumin If the area value is 0, it is marked as 1.

[0063] (4) Rock abundance

[0064] Rocks are one of the main features of the lunar surface and can pose a potential hazard to landers and rovers. It is defined as the proportion of the cumulative area covered by rocks with a diameter greater than or equal to a specific value (usually 10 cm) in a certain area, which can be obtained based on rock abundance image analysis.

[0065] When the abundance is greater than the threshold, the rock abundance accessibility map M rock The area value is marked as 0, indicating that it is inaccessible; when the abundance is less than the threshold, it is marked as 1.

[0066] Based on the feasibility analysis of the above four factors, four corresponding feasibility maps are obtained.

[0067] Through the four accessibility maps M slope 、M rough 、M illumin and M rock Perform intersection operation to obtain the final accessibility map M traverse That is, only the pixels that meet the traffic conditions under the four factors are marked as map M. traverse The formula is as follows.

[0068] M traverse =M slope ∩M rough ∩M illumin ∩M rock #(5)

[0069] With reference to the current dynamic parameter constraints of the Yutu lunar rover, the threshold of the maximum slope is set to 20°, and the threshold of the roughness is set to Cellsize / 5. With reference to the recommendations of similar missions, the threshold of rock abundance is limited to 7%. After multiple experimental tests, the threshold of the average illumination rate in the Ocean of Storms and the Chang'e 4 landing area is limited to 44%, and the threshold of the average illumination rate in the Antarctic region is limited to 7%.

[0070] Due to the accessibility map M traverse The pixel value on the grid only reflects whether the current grid is an obstacle point, and cannot fully reflect the density of obstacles in the area around the grid. In addition, the traditional A* algorithm usually uses the path length as an evaluation index when planning the path, and does not consider the distance between the planned path and the obstacle point. The planned path is close to the obstacle, which increases the risk of the lunar rover when driving. Therefore, the present invention generates the above-mentioned accessibility map M traverse Based on this, a 7x7 convolution kernel K is defined to calculate the safety of pixel points. The closer the point is to the center of the convolution kernel, the higher the weight. The convolution kernel K is as follows: Figure 2 .

[0071] For accessibility map M traverse For each pixel on the map, the following equations (6) and (7) are performed. If the pixel is an obstacle point (value is 0), the operation is skipped. Finally, a safety map M with a value range of 0 to 1 can be obtained. safety , where the larger the value, the higher the security of the point and the fewer obstacles around it; the smaller the value, the lower the security of the point and the more obstacles around it. When the value is 0, it means that the point is an obstacle and is not passable. The security map M after convolution calculation safety The pixel value on the grid can well reflect the density of obstacle points in the grid and its surrounding areas.

[0072]

[0073] Among them, M traverse (i,j) represents the map M traverse The pixel point (i, j) on the image is K(u, v), which represents the element in the convolution kernel K. Safety (i, j) represents the convolution result of the convolution kernel K at the pixel point (i, j), S K Represents the weight sum of the convolution kernel K.

[0074] Step 2: Generate a global planning path using a distributed computing strategy

[0075] S21. Improved A* algorithm (OC-WHT-A*)

[0076] The A* algorithm determines the priority of nodes by calculating the evaluation function F(N) = G(N) + H(N) of each node, where G(N) is the actual cost from the starting point to the current node N, and H(N) is the heuristic estimated cost from the current node to the target node. Hong et al. proposed an improved A* algorithm (OC-RA-A*Algorithm) with a random access data structure open-closed table, which uses a minimum heap and a two-dimensional array to jointly implement the open table, reducing the time complexity of determining whether an adjacent node is in the open table from O(n) to O(1). However, this method requires a large amount of memory space to be allocated for the two-dimensional array in advance when processing large-size images. In addition, the nodes actually visited by the A* algorithm are usually concentrated near the path from the starting point to the end point, which means that even if a very large two-dimensional array is allocated, only a small part of the space is actually used, resulting in a large waste of memory resources.

[0077] The present invention makes two improvements on the basis of the OC-RA-A* algorithm: first, the data structure of the open table and the close table of the algorithm is improved; second, the security of the path points is incorporated into the calculation of the estimated cost of the A* algorithm.

[0078] First, by improving the data structure, such as Figure 4 As shown in , the space complexity of the algorithm is significantly reduced while keeping the average time complexity of searching for nodes in the open table unchanged. Figure 4 As shown in the figure, in the improved A* algorithm based on the open and closed table data structure (OC-WHT-A*Algorithm), the data structure of the open table is implemented using a minimum heap and a hash table, and the data structure of the close table is implemented using a hash table. Compared with OC-RA-A*, the average time complexity of OC-HT-A* to determine whether the adjacent nodes of the current node are in the open table is also O(1), but the space complexity is increased from O(m 2 ) is reduced to O(n), where m is the size of the image and n is the number of nodes visited during the pathfinding process. The A* algorithm with improved data structure effectively reduces the space complexity while ensuring the search efficiency.

[0079] Second, in response to the problem of the traditional A* algorithm's path being close to obstacles, on the basis of the above improvements, the security of the path nodes is included in the calculation of the estimated cost. The lower the security of the path node N, the greater the estimated cost of the point. In this way, when searching for a path, the A* algorithm will give priority to nodes with higher security, avoid areas with more obstacles, and improve the security of the entire path.

[0080] The cost function calculation formula of the improved A* algorithm (OC-WHT-A*) is as follows:

[0081] F(N)=G(N)+(1.5-w×M Safety (N))×H(N)#(8)

[0082] F(N) represents the total cost from the starting point to the end point, G(N) represents the actual cost from the starting point to the current node N, H(N) represents the estimated cost from the current point N to the end point, w is the static weight, M Safety (N) is the accessibility of the current node N. When the accessibility of the path node is high, the coefficient before H(N) is smaller, and the total cost of the path point F(N) is also smaller. In addition, different sizes of w make the algorithm have different path selection tendencies. The larger w is, the slower the algorithm speed is, and it tends to find a path with higher security; the smaller w is, the faster the algorithm speed is, and it tends to find a path with a starting point and an end point first. The influence of weight w on the path is as follows: Figure 5 shown.

[0083] S22, global planning path generation

[0084] The present invention proposes a distributed path planning strategy DPPS-STP based on a security map tile pyramid, and implements it using a distributed computing engine Spark. The core process is as follows: first, read the top-level pyramid tile data from HDFS, construct a tile RDD, and filter related tiles according to the input start and end points. Then, perform distributed path planning on the filtered tiles to generate the path RDD of the layer. For non-bottom-level tiles, use the generated path RDD to infer the start and end points of the lower-level path planning, continue path planning, and complete the processing of the bottom-level tiles. Finally, write the path results into HDFS. The path planning on each layer of tiles is performed using the improved A* algorithm described below.

[0085] Using the properties of the pyramid model, the starting and ending points of each tile are calculated from coarse to fine. First, coarse-grained path planning is performed from the top tile layer, and the starting and ending points of the next tile layer are calculated based on the path results. Subsequently, path planning is performed on the next tile layer, and the divide-and-conquer idea is used to accelerate the sub-path planning tasks on each tile, and so on, until the path of the bottom tile layer is calculated, and finally the overall path planning result is obtained.

[0086] The distributed computing engine Spark is used to read the tile data blocks stored in HDFS layer by layer from top to bottom according to the pyramid level, and Spark's tile RDD is constructed for the tiles at each layer of the pyramid. The tile RDD is then distributed batch processed. During the batch processing, path planning is performed for each tile and path vertex RDD is generated. The above operations are iterated in Spark until the bottom-level path vertex RDD is obtained. Finally, the path vertex RDD is checked whether it is passable, and the path search of the global lunar grid DEM is re-performed for the inaccessible vertices in the path to ensure the accessibility of the search path.

[0087] In a master-slave cluster, Spark divides the path planning task into multiple subtasks and assigns them to slave nodes for execution. The slave nodes obtain tile data from HDFS, execute the subtasks assigned by the master node, and finally write the planning results to HDFS.

[0088] A schematic diagram of the iterative path planning process of a three-layer pyramid is shown in Figure 6 As shown. The red dots in the figure are the starting and ending points of the overall path of each layer of the tile pyramid, and the green dots are the points in the upper path points that are mapped on the boundary of the lower tile, that is, the starting and ending points of the sub-path planning tasks on each tile in each layer. First, according to the starting and ending coordinates of the top layer Layer-2, the rough path of the top layer of the pyramid is calculated, and then the path is mapped to the next layer of the pyramid Layer-1, that is, the path point coordinate value is multiplied by 2, and the point whose coordinate falls on the boundary of the lower tile is found. According to the coordinates of the path points on the boundary, the tiles containing these coordinates are filtered out, the path results are divided into multiple segments, and the sub-path results are calculated on each tile. This process is executed in parallel based on Spark. Repeat the above operations until the path result on the bottom layer Layer-0 is obtained, which is the final result.

[0089] Step 3: Careful cycle planning

[0090] The resolution of the full-moon DEM image is relatively low (20 meters / pixel). Although it can provide terrain information over a large range, it cannot provide sufficient details in local areas (especially areas with dense small craters). This results in the global path planning based on the full-moon DEM image possibly being unable to identify and avoid some small obstacles, thereby increasing the driving risk of the patrol vehicle. In order to solve this problem, the present invention introduces high-resolution DOM images (7 meters / pixel) to provide more detailed terrain information. The fine cycle planning proposed in the present invention mainly includes the following steps:

[0091] (1) Node simplification of the global planning path: The path nodes are simplified based on the Bresenham algorithm, and the simplified path nodes serve as the input for subsequent refined planning.

[0092] Since the 8-direction A* path planning algorithm is adopted, the planned path has many turning points, and the path turning angle is large and not smooth enough. The present invention optimizes the above path based on the Bresenham algorithm, deletes redundant path nodes, reduces the turning angle of the overall path, and finally obtains a group of optimized path nodes without obstacles between each other, which is used for more refined cycle planning.

[0093] The core idea of ​​the Bresenham algorithm is to use the slope of the line to advance the drawing pixel point by point, and select the discrete pixel point closest to the target line by incrementally calculating the decision variable. During the drawing process, the algorithm first determines the main direction of the line (i.e., the x-axis when the absolute value of the slope is less than 1, and the y-axis when it is greater than 1), and gradually advances with the main direction as the increment. Then, by calculating the vertical distance error from the current pixel to the line, the decision variable is updated to determine the position of the next pixel.

[0094] The core process of optimizing the path based on the Bresenham algorithm is as follows:

[0095] Starting from the starting point, traverse each path node in the path in turn, and determine whether the line between the current starting point and the traversal node passes through or is close to an obstacle point, that is, determine whether there is a dangerous node on the line. If not, continue traversing; if so, add the previous path node of the traversal node to the key point list, and set the previous path node as the starting point of the next round of traversal. Repeat this process until the end point is traversed, and finally generate a simplified path. Since only the path points on the line that do not contain dangerous nodes are added to the key point list, the simplified path can maintain the same safety as the original path.

[0096] The schematic diagram of the simplified path based on the Bresenham algorithm is as follows Figure 7 As shown in the figure, the black area is the obstacle area, the dotted part is the original path, and the solid part is the simplified path. Taking this figure as an example, first start from the starting point n0, and determine whether there is a dangerous node on the line between n0 and n1 based on the Bresenham algorithm. If there is no dangerous node, continue to traverse. When traversing to point n3, it is found that the line between n0 and n3 will pass through the obstacle point, so the previous node n2 of n3 is added to the key point list, and n2 is set as the starting point of the next round of traversal, and then start from point n2 and traverse backwards in sequence. Similarly, when traversing to point n5, it is found that there is an obstacle point on the line between point n2 and point n5, so the previous node n4 of n5 is added to the key point list, and the above process is repeated until the end point n is traversed. k , the path simplification process ends, and the points in the key point list are n0, n2, n4, n k, these points constitute the simplified path. Compared with the previous path, the simplified path shortens the path length and reduces the steering angle.

[0097] (2) Determine the scope of cycle path planning: Take every two adjacent nodes in the simplified path as a cycle, determine the longitude and latitude range of each path segment, and extract and crop the high-resolution DOM image of the corresponding area by calculating the geographic coordinates between the nodes.

[0098] (3) Small crater detection: The CenterNet network model based on deep learning is used to detect small craters in the extracted high-resolution DOM images. CenterNet is an efficient target detection algorithm that can accurately identify crater targets in images.

[0099] CenterNet is a deep learning-based object detection algorithm and one of the representatives of Anchor-Free methods. Its core idea is to simplify the object detection task into a key point detection problem. Traditional object detection methods usually rely on predefined candidate boxes (Anchors) to predict the position and size of the object by regressing these boxes. However, this method is not only computationally complex, but also easily affected by parameter settings and box position selection. In contrast, CenterNet abandons the Anchor mechanism and directly detects the center point of each object in the image, and infers the size, posture and other attributes of the object based on this, significantly improving the speed and accuracy while ensuring the detection speed.

[0100] The present invention uses CenterNet to detect craters. By superimposing features at different levels, more representative and robust hierarchical features are generated, which improves the ability to estimate the center of craters, especially the center of small craters. The smallest crater detected by this method is 500m. Compared with the Robbins crater database, the recall rate of crater detection by this method is 73.66% and the accuracy rate is 78.27%. The algorithm is publicly available at https: / / github.com / ShuoweiZhang / crater_detection.

[0101] (4) Safety map construction: The detected small craters are treated as obstacles and superimposed on the safety map described above. The safety map comprehensively considers factors such as terrain slope and crater distribution to provide support for local path planning. The OC-WHT-A* algorithm described above is re-adopted to generate a local path that avoids craters, thus achieving fine-grained periodic planning.

[0102] Figure 8The figure shows the comparison of path planning results before and after adding crater obstacles, where the red circle is the detected crater obstacle. It can be seen that most small craters are successfully identified. The blue path is the original path, that is, the path planning result without adding crater obstacles, and the green path is the result of periodic path planning after adding crater obstacles. It can be seen that the original path still passes through obvious obstacle areas (such as craters) on high-resolution DOM images, but by using CenterNet to detect small craters and adding obstacles to the safety map, the path can avoid obstacles that cannot be identified by DEM images alone, improving the safety of path planning.

[0103] To prove the effect of the technology proposed by the present invention, we conducted the following experiments:

[0104] 1. Evaluation indicators

[0105] This study uses time cost to measure the performance of the path planning algorithm. The time cost is defined as the total running time of the path planning. The time cost of the distributed algorithm includes the cluster startup time and the parallel computing time. The calculation formula is as follows:

[0106] T=T start +T compute #(9)

[0107] In addition, the path length, the number of dangerous nodes on the path, and the total turning angle of the path are used to evaluate the quality of the path results. The path length refers to the actual distance along the path from the starting point to the end point. A dangerous node is defined as a path node with an obstacle point within 40m around it. The total turning angle of the path is obtained by calculating the angle between the direction vectors of every three consecutive points on the path and gradually accumulating them.

[0108] 2. Experimental environment

[0109] All experiments were conducted in a Linux virtual machine environment with three 6-core, 16GB RAM, and CentOS 7 system built on a computer with Intel(R) Core(TM) i9-10920X CPU@3.50GHz and 64GB RAM.

[0110] First, a 4000x4000 area was selected in the Oceanus Procellarum region of the Moon for a comparative experiment in a single-machine environment. The improved algorithm was compared with the algorithm proposed by Hong et al., and the traditional A* algorithm was compared with the OC-HT-A* algorithm (without weight w) and the OC-WHT-A* algorithm (with weight w) proposed in this invention. The results are shown in Table 2 and Fig. 9 As shown, in Fig. 9In the figure, (a) path comparison of different A* algorithms under short distance, (b) path comparison of different A* algorithms under medium distance, and (c) path comparison of different A* algorithms under long distance.

[0111] Table 2 Comparison of short, medium and long distance path planning results of different A* algorithms in a single machine environment

[0112]

[0113] In the case of short distance, the traditional A* algorithm takes the longest time, which is 81.348 seconds; OC-RA-A* and OC-HT-A* take almost the same time, with a difference of less than 0.1 seconds. It can be seen that OC-HT-A* reduces the algorithm space complexity while not increasing the algorithm time complexity. OC-WHT-A* is the fastest, which is 6.93% higher than OC-HT-A*. The time cost of the traditional A* algorithm is about 10.96 times that of OC-HT-A* and about 11.78 times that of OC-WHT-A*. Since OC-RA-A* and OC-HT-A* only make improvements in data structure, their cost function design is the same as that of the traditional A* algorithm, so the path results obtained by the three are also the same, and the number of dangerous nodes is also the same, all 70. OC-WHT-A* improves the cost function H(N) and introduces the weight w to avoid areas with dense obstacles, so the number of dangerous nodes in the planned path results is significantly reduced to 0.

[0114] In the case of medium distance, the time consumption of the traditional A* algorithm increases significantly, which is 1418.512 seconds. The time consumption of OC-RA-A* and OC-HT-A* is similar. The OC-WHT-A* algorithm is the fastest, with a significant speed advantage over OC-HT-A*, which is improved by 48.74%. The time cost of the traditional A* algorithm is about 113.50 times that of OC-WHT-A*. It can be seen that as the distance increases, the path search space becomes larger, and the planning time of the traditional A* algorithm increases exponentially. OC-WHT-A* introduces the weight w, which optimizes the path search space, making it more efficient when planning a large-scale path. At the same time, the path planned by OC-WHT-A* still guarantees safety, and the number of dangerous nodes on its path is 0, which is significantly better than the path results planned by the other three algorithms.

[0115] In the case of long distance, the traditional algorithm takes 10145.400 seconds, while OC-RA-A* and OC-HT-A* take similar time. OC-HT-A* is slightly faster, taking 77.760 seconds. OC-WHT-A* is still the fastest, taking only 36.160 seconds, which is 53.49% faster than OC-HT-A*. In addition, the path planned by the OC-WHT-A* algorithm still ensures safety well, and the number of dangerous nodes on its path is 0. It can be seen that the traditional A* algorithm is difficult to apply to long-distance path planning, while the OC-WHT-A* algorithm can maintain the fastest planning speed in short, medium and long distances, and its performance advantage is constantly expanding as the distance increases. While ensuring speed, the OC-WHT-A* algorithm also ensures the safety of the path results, and avoids obstacles well in short, medium and long distances.

[0116] In addition, to verify the efficiency of the DPPS-STP distributed algorithm, the present invention selected three areas, namely, the Ocean of Storms, the Chang'e 4 landing area, and the lunar South Pole, for path planning experiments. The stand-alone OC-WHT-A* algorithm, the DPPS-STP algorithm using OC-HT-A*, and the DPPS-STP algorithm using OC-WHT-A* were tested and compared. The results are shown in Tables 3 and 4 and Figure 10-11 As shown, in Fig.11 In the figure, (a) shows the path comparison between the single-machine and distributed A* algorithms in the Ocean of Storms region, (b) shows the path comparison between the single-machine and distributed A* algorithms in the CE-4 landing area, (c) shows the path comparison between the single-machine and distributed A* algorithms in the Antarctic region, (d) shows an enlarged view of a local area in (a), (e) shows an enlarged view of a local area in (b), and (f) shows an enlarged view of a local area in (c).

[0117] In terms of time, it can be seen that the distributed algorithm significantly shortens the path planning time. The time of the DPPS-STP distributed algorithm using OC-HT-A* in the three areas is significantly lower than that of the single-machine OC-WHT-A* algorithm, with a maximum reduction of 90.15%, indicating the efficiency advantage of the DPPS-STP distributed algorithm in path planning. In addition, in the three areas, the time of the DPPS-STP distributed algorithm using OC-WHT-A* is lower than that of the DPPS-STP distributed algorithm using OC-HT-A*, and the time cost is reduced by an average of about 28.5%. This is because in the path planning calculation of a single tile, the DPPS-STP algorithm using OC-WHT-A* becomes faster due to the addition of weight w.

[0118] In terms of path length, the path lengths planned by the DPPS-STP distributed algorithm using OC-HT-A* in the three regions are all the shortest. The path lengths planned by the standalone OC-WHT-A* algorithm and the DPPS-STP distributed algorithm using OC-WHT-A* are similar. Compared with the DPPS-STP distributed algorithm using OC-HT-A*, the DPPS-STP distributed algorithm using OC-WHT-A* has a path length of 30.749 km longer in the Ocean of Storms, 23.96 km longer in the Chang'e 4 landing area, and 23.463 km longer in the Antarctic. The maximum increase in path length in the three regions is less than 8%.

[0119] In terms of the number of dangerous nodes, although the DPPS-STP algorithm using OC-HT-A* plans a shorter path than the DPPS-STP algorithm using OC-WHT-A*, the number of dangerous nodes on its path has greatly increased. The number of dangerous nodes in the path planned by the DPPS-STP algorithm using OC-HT-A* in the Ocean of Storms region is 604, and the number of dangerous nodes in the Chang'e 4 landing area and the Antarctic region is more than 1,000. However, the number of dangerous nodes in the path results of the DPPS-STP algorithm using OC-WHT-A* in the Ocean of Storms region and the Chang'e 4 landing area is 0, and the number of dangerous nodes in the path results in the Antarctic region is only 40, indicating that the path avoids the area around obstacles very well, ensuring the safety of the path results.

[0120] Table 3 Comparison of long-distance path planning time in stand-alone and distributed environments in three study areas

[0121]

[0122] Table 4 Comparison of long-distance path planning time, path length, and number of dangerous nodes in stand-alone and distributed environments in the three study areas

[0123]

[0124] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A lunar long-distance path planning method integrating terrain, illumination, and abundance characteristics, characterized by: First, a safety map is obtained by comprehensive analysis of DEM images, average illumination images and rock abundance images using safety evaluation rules. Then the safety map is constructed into a tile pyramid, divided into tiles by layer and stored in HDFS. The Spark distributed computing engine is then used to perform distributed computing on each layer of tiles from top to bottom to obtain a global planning path. Finally, combined with the DOM image, detailed periodic planning of the global planning path is implemented.

2. The lunar long-distance path planning method integrating terrain, illumination, and abundance characteristics according to claim 1 is characterized in that: When using the safety evaluation rules for comprehensive analysis, firstly, the DEM image, the average illumination rate image and the rock abundance image are analyzed for their accessibility, and four accessibility maps are generated accordingly. Then, the four accessibility maps are intersected to obtain a accessibility map. Finally, the accessibility map is convoluted to obtain a safety map. The accessibility map has only two values: 0 and 1, where 0 means that the point is an obstacle point and is not accessible; 1 means that this point is passable.

3. The lunar long-distance path planning method integrating terrain, illumination, and abundance characteristics according to claim 2 is characterized in that: Calculate the slope angle θ and roughness δ of each pixel in the DEM image, the average illumination rate μ of each pixel in the average illumination rate image, and the rock abundance of each pixel in the rock abundance image. Four accessibility maps were generated according to the following conditions; When the slope angle θ is greater than the first threshold, it is considered impassable and its value is 0; when it is less than the first threshold, its value is 1, so as to generate the slope passability map M slope ; When the roughness δ is greater than the second threshold, it is considered impassable and its value is 0; when it is less than the second threshold, its value is 1, so as to generate a roughness passability map M rough ; When the average illumination rate μ is less than the first threshold, it is considered impassable and its value is 0; when it is greater than the first threshold, its value is 1 to generate the illumination accessibility map M illumin ; When the rock abundance φ is greater than the first threshold, it is considered impassable and its value is 0; when it is less than the first threshold, its value is 1, so as to generate the rock abundance passability map M rock .

4. The lunar long-distance path planning method integrating terrain, illumination, and abundance characteristics according to claim 3 is characterized by: According to the following formula, the slope accessibility map M slope , Roughness passability map M rough , Light accessibility map M illumin , Rock abundance accessibility map M rock Perform intersection operation to obtain a traversability map M traverse ; M traverse =M slope ∩M rough ∩M illumin ∩M rock According to the following formula, the accessibility map M traverse Perform convolution calculation to obtain the security map M safety , Among them, M traverse (i,j) represents the accessibility map M traverse The pixel point (i, j) on the image is K(u, v), which represents the element in the convolution kernel K. safety (i, j) represents the convolution result of the convolution kernel K at the pixel point (i, j).

5. The lunar long-distance path planning method integrating terrain, illumination, and abundance characteristics according to claim 1 is characterized in that: The A* algorithm is improved from two aspects: cost function and data structure, and then the improved A* algorithm is used to perform path planning on each layer of tiles.

6. The lunar long-distance path planning method integrating terrain, illumination, and abundance characteristics according to claim 5 is characterized by: Use the following equation to calculate the cost function F(N) of the improved A* algorithm: F(N)=G(N)+(1.5-w×M safety (N))×H(N) Among them, F(N) represents the total cost from the starting point to the end point, G(N) represents the actual cost from the starting point to the current node N, H(N) represents the estimated cost from the current node N to the end point, w represents the static weight, M safety (N) represents the accessibility of the current point N in the safety map.

7. The lunar long-distance path planning method integrating terrain, illumination, and abundance characteristics according to claim 5 is characterized by: The data structure of the open table in the improved A* algorithm is implemented using a minimum heap and a hash table, and the data structure of the close table is implemented using a hash table.

8. The lunar long-distance path planning method integrating terrain, illumination, and abundance characteristics according to claim 5 is characterized in that Careful cycle planning includes the following steps: S1. Use the Bresenham algorithm to simplify the path nodes on the global planning path to obtain a simplified path; S2. For the simplified path, take every two adjacent path nodes as a cycle, determine the longitude and latitude range of each path segment, and extract and crop the high-resolution DOM image and security map of the corresponding area; S3, using the CenterNet network model based on deep learning to detect small craters on the extracted high-resolution DOM images; S4. The detected small craters are treated as obstacles and superimposed on the extracted safety map, and then the improved A* algorithm is used to re-plan the path to achieve cycle planning.

9. The lunar long-distance path planning method integrating terrain, illumination, and abundance characteristics according to claim 8 is characterized by: When the Bresenham algorithm is used to simplify the path nodes on the global planning path, starting from the starting point, each path node in the path is traversed in turn to determine whether the line between the current starting point and the traversal node passes through or is close to the obstacle point. If not, continue the traversal; if so, add the previous path node of the traversal node to the key point list, and set the previous path node as the starting point of the next round of traversal. Repeat the above process until the end point is traversed, and finally achieve path simplification.

Images