Towards Adaptive Interaction Strategies for Human Companion Robot via Deep Reinforcement Learning

This arXiv paper presents a novel approach to human-robot interaction in companion scenarios using deep reinforcement learning. Traditional methods constrain robots to fixed relative positions like tracking from behind, front, or side, which limits adaptability in dynamic workspaces. The study introduces adaptive interaction strategies that enable companion robots to flexibly adjust their positioning relative to humans in real-world environments. This work addresses practical challenges in service robotics and human-assistive applications where spatial interaction requirements constantly change.