Abstract
Automated vehicles are deemed to be the key element for the intelligenttransportation system in the future. Many studies have been made to improve theAutomated vehicles' ability of environment recognition and vehicle control,while the attention paid to decision making is not enough though the decisionalgorithms so far are very preliminary. Therefore, a framework of thedecision-making training and learning is put forward in this paper. It consistsof two parts: the deep reinforcement learning training program and thehigh-fidelity virtual simulation environment. Then the basic microscopicbehavior, car-following, is trained within this framework. In addition,theoretical analysis and experiments were conducted on setting reward functionfor accelerating training using deep reinforcement learning. The results showthat on the premise of driving comfort, the efficiency of the trained Automatedvehicle increases 7.9% compared to the classical traffic model, intelligentdriver model. Later on, on a more complex three-lane section, we trained theintegrated model combines both car-following and lane-changing behavior, theaverage speed further grows 2.4%. It indicates that our framework is effectivefor Automated vehicle's decision-making learning.