2018 年 54 巻 11 号 p. 793-801
Currently, vehicle's communication infrastructures are conducted in several loations, and they provide some traffic and signal information for vehicles. Using this information, automatic driving technologies are widely studied, and they focus on the avoidances of stopping by red signal and car crashes. They have deep insight for each research fields, however, there are no proposal for realizing both avoidance at same time. In this paper, we propose action determination learning scheme for realizing driving support considering both avoidance. Our scheme is based on Deep Reinforcement Learning similar with latest crash avoidance scheme. From simulation evaluation, the numbers of both crashes and stopping by red signal becomes lower by time goes on, and effective parameter tuning is revealed at same time.