Research On Autonomous Exploration Of Large-Scale 3D Environments And Application Of Search And Rescue

Autonomous exploration of mobile robots is the behavior of a robot moving in an unknown area and collecting information about its surroundings at the same time.After decades of development,autonomous exploration has gradually migrated from small-scale 2D indoor environments to large-scale 3D complex environments.Compared with small-scale 2D indoor environments,autonomous exploration in large-scale 3D environments is difficult to extract the required information from LiDAR native point cloud data,and has a higher degree of unknowns and dynamics.Search and rescue is one application of autonomous exploration in large-scale 3D environments.If time-sensitive search and rescue tasks are taken into account,the exploration efficiency of autonomous exploration needs to be emphasized.In this paper,we conduct a study on autonomous exploration technology for ground mobile robots in large-scale 3D environments and its application in search and rescue,with the following main work:1.A viewpoint-based accurate identification scheme is designed for the problems of missing partial point clouds,frontier omission and partial information gain miscalculation in the current unexplored area identification scheme.First,the LiDAR data is processed and geometric spatial features such as vertical surfaces and vertical frontiers are extracted.Then,the frontier viewpoints are extracted from the local viewpoints as virtual frontiers to enhance the guidance to unexplored areas.Finally,the problem of viewpoint gain miscalculation is corrected by using a normal filtering method.Through simulation experiments,the viewpoint gain miscalculation problem and the effectiveness of the normal filtering are verified,and the enhancement effect of the virtual frontier is also tested in publicly available large-scale 3D simulation environments.2.An efficient exploration strategy based on outside-first search is proposed for the short-sightedness problem of the traditional local optimal strategy in large-scale environments.First,a local planning algorithm is designed for the exploration of the environment.Then,for the zig-zag moving problem that occurs when the robot explores inward,a corresponding algorithm is given to optimize the selection of local goal viewpoints.Finally,a global planning algorithm is designed to ensure that the robot relocates to the remaining unexplored area after a local exploration is completed.The efficiency and real-time performance of the algorithm is verified by comparing it with two state-of-the-art algorithms in publicly available large-scale 3D simulation environments.3.For the search and rescue application requirements,an autonomous search and rescue system is designed based on the previously studied autonomous exploration system.First,the task requirements of search and rescue are analyzed,the functions of the autonomous search and rescue system are defined,and the system framework is designed.Then,the path planning algorithm for fast detection in the multi-victim’s scenario is given.Through simulation experiments,the efficiency of the algorithm and the feasibility of the autonomous search and rescue system are verified.