BatSLAM 2.0 is a sonar-only SLAM system for a robot carrying a biomimetic binaural sonar, modelled on echolocating bats. Sonar place recognition is ambiguous (corridors return nearly identical echoes) and one wrong loop closure can collapse a topological map, so the system adds a sequence verifier that tracks and checks loop-closure candidates before they enter a pose graph on a factor-graph back end. In simulation and real recordings it builds topological maps without collapse as the map grows.
Relevance
The sonar is in air, but the failure mode (ambiguous acoustic place recognition producing false loop closures) is the one underwater sonar SLAM also faces. Relevant to underwater perception and robust estimation in multi-domain perception.
Links
- arXiv
- Subfield: Mapping and localisation, Field robotics