On benchmarking of frontier-based multi-robot exploration strategies

Abstract

This paper tackles an evaluation and comparison of frontier-based exploration strategies to create a grid map of unknown environment by a team of autonomous mobile robots. A strategy is considered as a set of procedures to determine promising goal candidates, allocate them to the robots, and select the next navigational goal for each robot. A mobile robot is a complex system with many components that affect the mission performance and a comparison of different strategies in real experiments can be performed only for a particular system setup and with only partial controllability of important parameters. Therefore, the reproducibility and repeatability of such a comparison are not satisfied. In this paper, we propose a methodology for evaluating exploration strategies and provide a benchmark for a comparison of frontier-based approaches in a well-defined evaluation environment. The proposed methodology is demonstrated on a comparison of five state-of-the-art task-allocation strategies in multi-robot exploration.