Sample-Based Frontier Detection for Autonomous Robot Exploration

Frontier-based method is most commonly used in robotic exploration. One popular frontier searching method is to exploit the idea of rapidly-exploring random tree and to use the grown edges of the tree to search for frontiers. Compared to traditional methods based on image processing, it can be applied to high-dimensional exploration more efficiently. However, this method usually needs to occupy a large number of storage resources and searches for frontiers slowly in the environment where random trees are not easy to grow (unfavorable environment). In this paper, a sample-based frontier detection algorithm (SFD) is proposed. Firstly, by changing the growth rule and the storage mode of the random tree, the disadvantage of slow growth of the tree under unfavorable environments is overcome. Secondly, we divide the map into blocks which are used to delete redundant tree nodes during the exploration to reduce required computation resources. In order to evaluate the proposed frontier detection algorithm, two different kind of simulation environments have been set up. The experimental results show that our algorithm saves the memory resource greatly and shows better performances in unfavorable environments.