An Optimal Intersection Control Policy For Autonomous Traffic Management Using Multiagent Approach

Multiagent System (MAS) is one of the most important subfields of Artificial Intelligence (AI), which mainly aims to provide both principles for construction of complex systems involving multiple agents and mechanisms for coordination of independent agent’s behaviors. Currently on the basis of MAS, removing the human driver from the control loop through the integration of autonomous vehicles and intelligent road infrastructures has been a reality.In this paper, we have studied an autonomous traffic management mechanism for intersections using the multiagent approach. We propose a maximum-clique inspired optimal intersection control policy ConflictReduce (CR) with regard to the blindness and lack of foresight of the core control policy FCFS for the previous system. Through evaluation and validation on the open-source simulator AIM, our policy is shown to have significantly better performance in managing and scheduling the large-scale autonomous traffic in the intersection. Compared with FCFS policy and Traffic Light at the extreme traffic condition of2500vehicles per lane per hour, the average traffic delay for CR policy can still be considerably reduced by over40%and nearly16%respectively, and the traffic throughput can be dramatically improved by over30%and closely20%as well.In order to better cope with both of the issues in real-time property and communication complexity of the system due to the introduction of the new control policy, we further present the dynamic adjustment policy for batch processing cycle and the incremental transmission policy ksync for data synchronization. We indicate the remarkable achievement in the reduction of system communication complexity by empirical evaluation on AIM. The simulation result shows that the average data compression rate can be improved by over80%for the reserved driver agent exploiting ksync policy.