SAMALM is a novel framework that uses multiple LLM agents in an actor-critic structure to enable socially-aware robot navigation in dynamic human environments.

• Combines the reasoning capabilities of LLMs with reinforcement learning for more adaptive path planning
• Creates a multi-agent system where LLMs evaluate and improve robot navigation decisions
• Demonstrates superior adaptability to new scenarios compared to traditional deep reinforcement learning approaches
• Enhances robot safety and social awareness in human-populated spaces

This research bridges the gap between advanced AI language models and physical robotics, enabling more intuitive and safe human-robot interactions in everyday environments without requiring extensive retraining for new scenarios.

Multi-Agent LLM Actor-Critic Framework for Social Robot Navigation