Time-Efficient Autonomous Exploration in Unknown Environment by Multirepresentation Strategy

Abstract

Autonomous exploration of unmanned ground vehicles (UGVs) is a critical yet challenging task that has garnered widespread attention and experienced rapid progress. In exploration tasks, redundant revisits to known areas frequently occur, significantly impacting exploration efficiency. To address this issue, we introduce a multirepresentation exploration strategy with a long-short path approach. First, we use topological and terrain maps to represent local frontiers. Topological maps enhance the sensor’s spatial perception capabilities, while terrain maps provide detailed map expressions. Together, they improve exploration efficiency. Second, the long-short path strategy is applied to update the global topology, making its construction more robust. Third, our algorithm incorporates topological information into the decision-making process, enhancing both the efficiency and coherence of exploration. In our experimental evaluations, our algorithm was benchmarked against mainstream algorithms and demonstrated superior performance across various indicators. Compared to other algorithms, our algorithm is at least 4.54% faster in total time and reduces the total traveling distance by 4.82%.