Robot Exploration based on Small Areas Priority Strategy

Abstract

We present a method for robot exploration in complex three-dimensional environments. We prove that when facing multiple path branches, it is a necessary condition for optimal robot exploration to explore the branches connecting independent areas first, so we propose a local exploration strategy based on small areas priority. In order to realize this strategy, our method models lidar points and builds a prediction map covering more spatial information obtained by the sensor at the expense of accuracy. This method evaluates the potential explorable area of the unexplored space through ray-casting and the path tree, and decides which of the explorable path branches is the optimal choice in the local range. The method is compared to existing state-of-the-art methods in two corridor simulation scenes and three simulation scenes replicated from the real world. Experiment comparisons show that our method is more efficient in exploring space and closer to the global optimal trajectory.