2017/09/14 by De Silva, De Silva, Varuna, Xiongzhao Wang +7
Computer Science · Engineering · #Autonomous Vehicle Technology and Safety #FOS: Computer and information sciences #Multiagent Systems (cs.MA) #Reinforcement Learning in Robotics #Robotics (cs.RO) #Traffic control and management #Transportation and Mobility Innovations
paper · pdf · doi:10.48550/arxiv.1709.04622
openalex publication_date 2017/09/14 · openalex created_date 2022/10/02 · openalex updated_date 2026/07/28
Due to the complexity of the natural world, a programmer cannot foresee all\npossible situations, a connected and autonomous vehicle (CAV) will face during\nits operation, and hence, CAVs will need to learn to make decisions\nautonomously. Due to the sensing of its surroundings and information exchanged\nwith other vehicles and road infrastructure, a CAV will have access to large\namounts of useful data. While different control algorithms have been proposed\nfor CAVs, the benefits brought about by connectedness of autonomous vehicles to\nother vehicles and to the infrastructure, and its implications on policy\nlearning has not been investigated in literature. This paper investigates a\ndata driven driving policy learning framework through an agent-based modelling\napproaches. The contributions of the paper are two-fold. A dynamic programming\nframework is proposed for in-vehicle policy learning with and without\nconnectivity to neighboring vehicles. The simulation results indicate that\nwhile a CAV can learn to make autonomous decisions, vehicle-to-vehicle (V2V)\ncommunication of information improves this capability. Furthermore, to overcome\nthe limitations of sensing in a CAV, the paper proposes a novel concept for\ninfrastructure-led policy learning and communication with autonomous vehicles.\nIn infrastructure-led policy learning, road-side infrastructure senses and\ncaptures successful vehicle maneuvers and learns an optimal policy from those\ntemporal sequences, and when a vehicle approaches the road-side unit, the\npolicy is communicated to the CAV. Deep-imitation learning methodology is\nproposed to develop such an infrastructure-led policy learning framework.\n