You can go first - Planning for social law synthesis in asymmetric multi-agent settings

Sarath Sreedharan, Philip D. Hopkins, Felipe Meneguzzi

Keywords: Social Norms, Stackelberg Planning, human-robot interaction

Abstract: Social norms are a well-known mechanism to implement agent coordination without explicit communication, runtime negotiation, or centralized planning. While synthesizing social norms in general for a multi-agent system is a hard problem, we focus on a specialized problem where one of the agents accepts taking suboptimal actions, and lets the other act optimally, while avoiding potential conflicts. We show how to encode this problem into problems amenable to using existing planning systems as solvers. We prove that the resulting encodings are complete, guaranteeing us the ability to identify social norms that can support such coordination if one exists, and provide empirical results comparing these compilations.

Area: Search, Optimization, Planning, and Scheduling (SOPS)

Generative A I: I acknowledge that I have read and will follow this policy.

Submission Number: 1449