Connected autonomous vehicles for improving mixed traffic efficiency in unsignalized intersections with deep reinforcement learning

Human driven vehicles (HDVs) with selfish objectives cause low traffic efficiency in an un-signalized intersection. On the other hand, autonomous can overcome this inefficiency through perfect coordination. In paper, we propose intermediate solution, where use vehicular communication and a small number of to improve transportation system such intersections. our two connected (CAVs) lead multiple HDVs double-lane intersection order avoid congestion front The CAVs are able communicate coordinate their behavior, which is controlled by deep reinforcement learning (DRL) agent. We design altruistic reward function enables adjust velocities flexibly queuing proximal policy optimization (PPO) algorithm applied train generalized advantage estimation (GAE) used estimate state values. Training results show that achieve significantly better compared similar scenarios without one vehicle.