Multiagent Dynamic Route Assignment: Quick and Fair Itineraries to Connected and Autonomous Vehicles

Abstract

In recent years, Connected and Automated Vehicle (CAV) related research has progressed considerably. This paper proposes an approach for finding the best itinerary to CAVs into a road network. If we consider only a single CAV, the problem turns into the shortest path problem according to the real-time traffic information. When it comes to a road network served by several CAVs, such as in Personal Rapid Transit (PRT), the problem of the traffic assignment equilibrium is raised. However, the known algorithms for solving such problems are greedy in terms of computation time and memory. In order to solve the problem, this paper introduces a new distributed approach, using multi-agent systems. We call the approach Mixed Node Reservation (MNR). Each CAV agent computes its itinerary by using an accurate estimation of its travel time through its connectivity to node agents of the network. To this end, two new key concepts are introduced. The first one is the node reservation. This allows to adjust the travel time between two nodes, by considering the acceleration time as well as the time spent into the next intersection. The second key concept is the spawned virtual CAVs. This aims to accurately estimate the future lost time generated by CAVs which are not yet in the network. Other approaches of the literature are simulated for comparisons. The experiment results show that MNR allows CAVs to share the network much more efficiently.