关键词: 未知环境探索, 自适应采样框, 重要性采样, 路径规划, ＳＬＡＭ

Abstract:  Boundary sensing detection is one of the important parts of autonomous exploration of UAV. In order to improve the efficiency of boundary detection in the process of autonomous exploration in complex and diverse underground narrow environments, this paper proposes an adaptive fast boundary detection algorithm for unmanned aerial vehicles (ADPlanner) in unknown environments. First, we perceive the unknown environment of the underground tunnel through optical radar, adaptively adjust the local sampling space of the underground tunnel or mine tunnel environment, and greatly improve the sampling rate (the ratio of the sampling points added to the RRG to the number of sampling attempts) according to the environmental structure. Secondly, we propose a resampling rate to reduce the redundancy of sampling points of the adjacent adaptive sampling frame, Then, the importance sampling strategy is used to solve the oversampling problem of repeated areas in the GBPlanner and achieve incremental detection. The simulation experiment shows that in two different unknown scenarios, compared with the GBPlanner, the ADPlanner boundary detection sampling run time is reduced by 20.27%～38.33%, the path length is reduced by11.24%～18.86%, and the total exploration time is shortened by 27.38%～38.38%, which significantly improves the exploration efficiency of the UAV in the unknown environment.。