Autonomous Vehicles in Smart Cities: a Deep Reinforcement Learning Solution

In order to alleviate the traffic congestion problem in the context of Smart Cities, this paper proposes a new routing strategy, based on Deep Reinforcement Learning. The idea is to perform Routing actions in real-time while considering the actual traffic situation on the overall road network.To evaluate the performance of the proposed solution, a co-design procedure, involving a microscopic traffic simulator and MATLAB environment, is implemented allowing the use of complex operating scenarios and the integration of information coming from road maps and vehicle state trajectories.Finally, some simulations performed considering a real city district, show that the proposed algorithm can successfully realize real-time routing decisions and improve accumulated waiting time although the dynamic changes inside the road environment.