Multi Agent Based Traffic Controls

Abstract

The traffic jams are common in the modern world and are mostly attributed to increase in population and the continual expansion of the urban centers. Traffic jams cost the economy a lot of money in wasted time in traffic and fuel consumption and also result to air pollution. Advances in the field of artificial intelligence have made it suitable to use agents in the management and control of traffic due to their ability to control and coordinate their activities. This research project has implemented a multi agents based traffic control system that is able to manage the traffic flow based on the prevailing conditions on the roads. Multi agent systems are best suited for such environments since they are able to perceive the environment they are located in and make decisions accordingly by negotiating and cooperating to ensure smooth flow of traffic regardless of the traffic densities. The agents negotiate based on the average waiting time and queue lengths such that agents whose junction has a maximum waiting time and queue lengths are given preference. By use of a simulator comparisons in performance have been done between a pre-timed traffic control system and a multi agent based traffic control system. Results show that the performance of both systems deteriorates with increase in traffic volumes. Regardless of this a multi agent based traffic control system is able to perform better than the pre-timed traffic control system regardless of the traffic situation on the roads by attaining 33% improvement.