Multi-Agent Strategy Explanations for Human-Robot Collaboration

• DOI: 10.48550/arxiv.2311.11955
• Publication Date: 2023-01-01
• Authors: Ravi Pandya; Michelle Zhao; Changliu Liu; Reid Simmons; Henny Admoni
• Keywords: Planner; Landmark; Shared space; Human–robot interaction

As robots are deployed in human spaces, it's important that they able to coordinate their actions with the people around them. Part of such coordination involves ensuring have a good understanding how robot will act environment. This can be achieved through explanations robot's policy. Much prior work explainable AI and RL focuses on generating for single-agent policies, but little has been explored collaborative policies. In this work, we investigate generate multi-agent strategy human-robot collaboration. We formulate problem using generic planner, show visual strategy-conditioned landmark states textual by giving landmarks an LLM. Through user study, find when presented from our proposed framework, users better explore full space strategies collaborate more efficiently new partners.