Each robot builds a Gaussian-process occupancy map with a conformal coverage guarantee, but the guarantee degrades because samples along a trajectory are correlated and each robot sees only part of the area. This paper takes those degraded local guarantees as given and fuses them across the team: neighbours exchange only scalar e-values, and the receiver combines them with a per-neighbourhood miscoverage budget. The fused map provably recovers the target coverage for any communication graph and noise distribution, at the price of refusing to label cells where evidence is thin; in simulation, denser graphs shrink that unlabelled fraction.
Relevance
One of the few multi-robot mapping papers that states what guarantee the merged map keeps when each robot’s own guarantee is already broken. See multi-robot perception, robust robotics and collaborative SLAM. The abstract says nothing about faulty or lying neighbours, so what one bad e-value does to the fusion is an open question.
Links
- arXiv
- Subfield: Multi-robot systems, Mapping and localisation