Cooperative Decision Strategies Applied to a Distributed Sensor Network

This paper outlines the research conducted on the SEAS DTC AA011 project into learning cooperative decision strategies for Autonomous Vehicles (AVs). The paper presents a brief review of the technical challenges involved and existing approaches to designing cooperative decision strategies. An initial study is used to compare two of the candidate solutions: Probability Collectives (PC) and Coordinated Reinforcement Learning (CRL). The paper concludes with experimental results of using PC to enable cooperative behaviour in a distributed sensor network. The results show that the cooperation provided from using PC has significant improvements in performance over a purely selfish strategy.