Real-Time Autonomous Taxi Service: An Agent-Based Simulation

Negin Alisoltani, Mahdi Zargayouna, Ludovic Leclercq

Abstract: Today policymakers face increasingly complex traffic systems. While they need to ensure smooth traffic flows in the cities, they also have to provide an acceptable level of service in remote areas. Autonomous Taxis (AT) offer the opportunity to manage car traffic at low operational cost and they can be appropriate alternatives for the driven vehicles. In this paper, we propose a multi-agent system to find the best dispatching strategy for a fleet of AT. In the dispatching process, we aim to satisfy both the passengers and the providers objectives and priorities. To be able to apply the method on large-scale networks, we introduce a clustering method to cluster the requests every minute and then solve the assignment problem for the requests inside each cluster. As the network congestion can have significant impacts on the vehicle speed and travel time especially considering the private cars that are driving in the system besides the taxis, an agent-based simulation platform is used to simulate the function of the AT fleet. We use the trip-based Macroscopic Fundamental Diagram (MFD) to simulate the time evolution of traffic flows on the road network and update the traffic situation in the system every second to represent the real traffic dynamics. We address the problem for a large city scale of 80 km\(^2\) (Lyon city in France) with more than 480,000 trips over 4 h period containing the morning peak. The experimental results with real data show that the proposed multi-agent system is efficient in terms of serving all the requests in a short time satisfying both passengers and providers objectives.