Dylan P. Losey*, Mengxi Li*, Jeannette Bohg, Dorsa Sadigh

Abstract: When teams of robots collaborate to complete a task, communication is often necessary. Like humans, robot teammates should implicitly communicate through their actions: but interpreting our partner’s actions is typically difficult, since a given action may have many different underlying reasons. Here we propose an alternate approach: instead of not being able to infer whether an action is due to exploitation, information giving, or information gathering, we define separate roles for each agent. Because each role defines a distinct reason for acting (e.g., only exploit), teammates can now correctly interpret the meaning behind their partner’s actions. Our results suggest that leveraging and alternating roles leads to performance comparable to teams that explicitly exchange messages.

CoRL 2019 Conference Paper (oral) : available on arXiv

Blogpost: available on Stanford AI Lab Blog

Contact: dlosey {at} stanford {dot} edu, mengxili {at} stanford {dot} edu for more information

Bibtex:

@inproceedings{losey2019learning,

   title={Learning from My Partner's Actions: Roles in Decentralized Robot Teams},

   author={Losey, Dylan P and Li, Mengxi and Bohg, Jeannette and Sadigh, Dorsa},

   booktitle={Conference on Robot Learning},

   year={2019},

   url={https://arxiv.org/abs/1810.10191}

}

Research supported by: