Research On Lunar Path Planning Based On Remote Sensing Big Data Platform

In the nationally implemented lunar exploration project,lunar rovers play an important role in the exploration of the lunar environment by collecting data from the Moon so that humans can learn more about space.In lunar exploration missions,the route planning for the rover is the foundation of the lunar exploration project.In particular,planning the route of the rover to its destination quickly and safely can help people understand the topography,composition,and geological history of the lunar surface more efficiently and deeply,thus advancing the process of exploring space.In order to achieve fast,safe and efficient route planning,scholars have developed many advanced techniques and algorithms,such as Dijsktra algorithm,A* algorithm,genetic algorithm,simulated annealing algorithm and so on.These techniques can help people to generate the best route automatically after considering multiple factors,thus improving the efficiency of rovers for lunar exploration missions.With the growing development of modern mapping technology and sensor hardware,the Digital Elevation Model(DEM)has become more and more accurate,which makes path planning based on DEM a reality.In lunar exploration missions,DEM data can help scholars better understand the topographic features of the lunar surface and the distribution of mineral resources,so as to determine the route of the rover.However,with the accuracy of DEMs increasing,the research on lunar path planning faces new challenges.On the one hand,the increase of DEM data makes the increasing of data scale available for performing path planning,thus requiring the more efficient algorithms and techniques to perform path planning tasks.On the other hand,DEM-based path planning requires a lot of computation and data processing,especially in terms of computation time and memory usage,thus requiring the use of multi-server devices to improve computation and storage efficiency.In this paper,we focus on fast and safe lunar path planning for lunar rovers on lunar DEM data covering more than 42,996 square kilometers,and the main research is as follows:1.When performing path planning tasks on the lunar DEM,an A* pathfinding algorithm with improved data structures(OC-RA-A* algorithm)is proposed in this paper to address the problems of slow pathfinding speed and long pathfinding time generated with existing pathfinding algorithms.The efficiency of the A* pathfinding algorithm is improved by utilizing the in-memory characteristics of the data structure to improve the operation efficiency of the open queue Open-List and the queried queue Closed-List in the A* algorithm.Research experiments show that the OC-RA-A* algorithm is 3.59 times faster than the traditional A* algorithm in a long-distance pathfinding task of over 934 km.2.When performing the path planning task on the lunar DEM with large data scale,this paper proposes a distributed storage and computation method based on the tile pyramid of DEM for the problem of limited hardware resources when running the task on a single server.By utilizing Hadoop Distributed File System(HDFS)and Spark’s distributed computing engine,a remote sensing big data platform is built for the distributed tile pyramid storage and computation for DEM.3.Because of the limited computing and storage resources of the A* algorithm on the large-scale lunar DEM data when using single server,and the issues of unable to effectively utilize the neighborhood information of the DEM when using Spark in distributed environment to perform A* algorithm path planning of discrete points of the lunar DEM,this study proposes a tile pyramid-based distributed path planning strategy(DPPS-TP).By using the distributed computing power of Spark and the neighborhood information that can be preserved by the distributed tile pyramid model,DPPS-TP improves the time efficiency of path planning tasks on the lunar DEM.The distributed computing power of Spark ensures that DPPS-TP is not affected from the limited resources of single machine,and the distributed tile pyramid model can retain relatively more neighborhood information.The experimental results show that when using the OCRA-A* algorithm as the path planning algorithm,the long-distance pathfinding task based on DPPS-TP is 1.57 times faster than that based on the single-machine serial pathfinding task;when using the traditional A* algorithm as the path planning algorithm,the longdistance pathfinding task based on DPPS-TP is 4.34 times faster than that based on the single-machine serial pathfinding task.When using Spark as the parallel computing engine,the long-distance pathfinding task based on DPPS-TP is 113.66 times faster than the pathfinding task using only raster DEM discrete points.Moreover,the pathfinding task based on DPPS-TP is more stable in terms of time efficiency and does not lead to exponential increase in computation time when the data volume increases compared to the pathfinding task based on single machine serial and based on Spark parallelism using only raster DEM discrete points.