MA-CoNav: A Master-Slave Multi-Agent Framework with Hierarchical Collaboration and Dual-Level Reflection for Long-Horizon Embodied VLN

Researchers propose MA-CoNav, a multi-agent collaborative framework for robot navigation that uses a Master-Slave architecture to distribute cognitive tasks among specialized agents. The system outperforms existing Vision-Language Navigation methods by decoupling perception, planning, execution, and memory functions across different AI agents with hierarchical collaboration.

• →MA-CoNav introduces a Master-Slave multi-agent architecture that distributes navigation tasks among specialized agents to reduce cognitive overload.
• →The framework includes four specialized agents: Observation, Planning, Execution, and Memory agents, each handling distinct navigation functions.
• →A dual-stage Local-Global reflection mechanism dynamically optimizes the entire navigation pipeline for better performance.
• →Real-world testing on indoor datasets showed comprehensive performance improvements over existing VLN methods without scene-specific fine-tuning.
• →The approach addresses perceptual distortion and decision drift issues in complex, long-distance robotic navigation tasks.