Optimizing vehicle trips using agent negotiation through a traffic matrix

Abstract

Multi-Agent Systems (MAS) have been proved to be an effective tool to improve the efficiency of intelligent traffic control systems (using them in traffic lights scheduling, vehicle routing, etc.). Negotiation is one of the most used techniques to improve the global goal of a system while maintaining as much as possible each agent preferences. In this paper, we propose modeling the traffic characteristics produced in a transportation network during a given time interval as a traffic matrix, and then use it as core element in an agent negotiation system. We define how this matrix is generated, and the mechanisms to populate it and update it. Using the traffic matrix data, we propose to use two different selection methods to obtain a subset of agents with voting rights. The first method is based on the average speed of network edges and the second one is based on the traffic jam length for a vehicle. This subset will perform a negotiation process where other agents in the network will be proposed to block their departure or modify their departure time, to mitigate the congestion in the network edges, and as goal, reducing trip duration.