HEROIC coordinates a decentralised air-ground team for open-vocabulary search, with agents talking only in natural language. A scale law based on sensor properties sets drone altitude and sweep spacing from the mission prompt, and when the required altitude is below the safe flight height, drones switch from searching to triage, escort and route guidance for the ground robots. Each robot keeps an evidential belief over the area (bearing rays for positive evidence, a log-odds posterior for negative evidence) and confirms targets at close range; HEROIC reaches the target in 84% of runs across six scenes against 35 to 54% for the baselines, and 2 to 4 times sooner.
Relevance
Roles that change with what each platform can actually sense are the heterogeneous-team problem in small. Relevant to multi-robot multimodal perception and semantic coordination; language-only messages also raise the question of how a team checks a teammate’s claims.
Links
- arXiv
- Subfield: Multi-robot systems, Field robotics, Semantic mapping