Training-Free Multi-Agent Cooperative Embodied Search from Uncertain Vision-Language Priors

Key points

• This paper presents a lightweight, training-free, and uncertainty-aware framework for cooperative multi-agent multi-target search in mapless environments.1
• Instead of learning an end-to-end navigation policy, the framework separates high-level semantic reasoning from low-level execution through a scalable graph-optimization backbone.1
• Multimodal Large Language Models (MLLMs) serve only as hypothesis generators, proposing semantic zones, target-location cues, and possible navigable connections.1
• Cooperative navigation is then formulated as a rolling-horizon multi-agent mixed-integer linear program with a lexicographic objective that first maximizes target-directed reward and then reduces travel distance, uncertain-edge use, and repeated visits.1

Sources (1)

• [1]Training-Free Multi-Agent Cooperative Embodied Search from Uncertain Vision-Language Priors
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 10, 04:22 PM

  Real-world Vision-Language Navigation (VLN) requires agents to act under partial observability, where map structure, semantic cues, and target locations may be incomplete or incorrectly inferred by foundation models.

  This paper presents a lightweight, training-free, and uncertainty-aware framework for cooperative multi-agent multi-target search in mapless environments.

Extractive summary: sentences quoted from the sources.