Paper on Multi-Agent Reinforcement Learning accepted at IROS 2020!

Achieving shared goals with Stackelberg learning in cooperative control


by Alessandro Roncone on December 08, 2020

Congratulations to Joewie and Guohui for their paper “Cooperative Control of Mobile Robots with Stackelberg Learning”, which was published at IROS 2020!

The paper proposes Stackelberg Learning in Cooperative Control (SLiCC), a method for cooperative control of multi-agent systems in partially observable settings. SLiCC is based on an asymmetric prosocial–introspective cooperation framework that links state perception with agents’ decision-making strategies. This framework allows for agents to have different observation scopes, with prosocial and introspective behaviors assigned to agents based on the completeness of their state perception.

[PDF] [BIB]