Only content delivered through the publication boundary on this date is included.
arXiv This paper presents DM³-Nav, a fully decentralized multi-agent semantic navigation system supporting multimodal goal specification (category, language, and image) and multi-object episodes, where robots coordinate solely through ad-hoc pairwise communication exchanging local maps, goal status, and navigation intent without a central coordinator or shared global map; evaluations on HM3DSem scenes (HM3Dv0.2 and GOAT-Bench) show it matches or exceeds centralized and shared-map baselines, and a two-robot team successfully located all 8 multimodal targets in a real-world office environment.
This paper presents DM³-Nav, a fully decentralized multi-agent semantic navigation system supporting multimodal goal specification (category, language, and image) and multi-object episodes, where robots coordinate solely through ad-hoc pairwise communication exchanging local maps, goal status, and navigation intent without a central coordinator or shared global map; evaluations on HM3DSem scenes (HM3Dv0.2 and GOAT-Bench) show it matches or exceeds centralized and shared-map baselines, and a two-robot team successfully located all 8 multimodal targets in a real-world office environment.
This paper presents DM³-Nav, a fully decentralized multi-agent semantic navigation system supporting multimodal goal specification (category, language, and image) and multi-object episodes, where robots coordinate solely through ad-hoc pairwise communication exchanging local maps, goal status, and navigation intent without a central coordinator or shared global map; evaluations on HM3DSem scenes (HM3Dv0.2 and GOAT-Bench) show it matches or exceeds centralized and shared-map baselines, and a two-robot team successfully located all 8 multimodal targets in a real-world office environment.
This paper presents DM³-Nav, a fully decentralized multi-agent semantic navigation system supporting multimodal goal specification (category, language, and image) and multi-object episodes, where robots coordinate solely through ad-hoc pairwise communication exchanging local maps, goal status, and navigation intent without a central coordinator or shared global map; evaluations on HM3DSem scenes (HM3Dv0.2 and GOAT-Bench) show it matches or exceeds centralized and shared-map baselines, and a two-robot team successfully located all 8 multimodal targets in a real-world office environment.