Autonomous Exploration for Robot in Tunnel Scene Based on Low‐Credibility Prior Topological Map

ABSTRACT

Underground mines, natural caves and other tunnel‐like environments are typically long and narrow, often featuring numerous intersections and sharp turns. These features exacerbate the short‐sighted decisions of frontier‐based exploration methods, resulting in inefficient and redundant exploration. In search and rescue or geological exploration tasks, escape instructions and fire relief maps provide prior topological maps that describe the spatial layout of tunnels. These maps can help robots foresee tunnel layouts before entering. Because prior maps may be outdated or inaccurate due to environmental changes, an effective data fusion method is needed to combine prior topological maps and real‐time data to plan more effective exploration paths. This paper proposes an autonomous exploration method that utilises a low‐confidence prior topological map, enabling the robot to efficiently explore tunnels.