Minimizing Conflicts Between Moving Agents over a Set of Non-Homotopic Paths Through Regret Minimization
Title | Minimizing Conflicts Between Moving Agents over a Set of Non-Homotopic Paths Through Regret Minimization |
Publication Type | Report |
Year of Publication | 2013 |
Authors | Kimmel, A, Bekris, KE |
Date Published | 07/2013 |
Institution | AAAI 2013 Workshop on Intelligent Robotic Systems |
City | Bellevue, WA, USA |
Abstract | This paper considers a game-theoretic framework for motion coordination challenges. The focus of this work is to minimize the number of interactions agents have when moving through an environment. In particular, agents employ a replanning framework and regret minimization over a set of actions, which correspond to different homotopic paths. By associating a cost to each trajectory, a motion coordination game arises. Regret minimization is argued as an appropriate technique, as agents do not have any knowledge of other agents’ cost functions. This work outlines a methodology for minimizing the regret of actions in a computationally efficient way. Initial simulation results involving pairs of mobile agents show indications that the proposed framework can improve the choice of non-colliding paths compared to a greedy choice by the agents, without increasing any information requirements. |
URL | http://www.cs.rutgers.edu/~kb572/pubs/work_aaai_13_regret_minimization.pdf |
Refereed Designation | Refereed |