SALM framework combines Large Language Models with Deep Reinforcement Learning to enable robots to navigate human spaces while responding to real-time commands.

• Creates socially-aware navigation capabilities in zero-shot environments
• Enables interactive feedback from humans during navigation tasks
• Supports complex services like cooperative carrying in crowded spaces
• Ensures robots maintain socially compliant behaviors while navigating

This engineering breakthrough allows service robots to understand human intent, adapt to dynamic environments, and follow social norms—critical capabilities for next-generation interactive robotics solutions in public spaces.

Unifying Large Language Model and Deep Reinforcement Learning for Human-in-Loop Interactive Socially-aware Navigation