Decentralized intersection management through peer-to-peer technology

In this paper we present two new decentralized algorithms for autonomous intersection management and compare the performance of the algorithms with an established centralized solution. One of the algorithms addresses the problem through an unstructured peer-to-peer approach and the other uses a Distributed Hash Table to distribute knowledge of intersection usage among participating vehicles. We evaluate these algorithms through simulation and by comparing average delay to the performance of a centralized reservation-based algorithm. We find that in times of light congestion the decentralized algorithms perform as well as the centralized approach. During times of moderate or heavy congestion the unstructured peer-to-peer algorithm performs better than the centralized algorithm, and the DHT-based algorithm performs worse.