A Study on Community Road Safety Measures Using Inverse Reinforcement Learning

Abstract

The occurrence of traffic accidents in Japan has continued to decrease, but in recent years it has bottomed out. This trend is more pronounced on community roads. Compared to highway, it has been difficult to obtain data on traffic volume and other data on community roads, limiting the analysis of safety measures. In recent years, however, the availability of data on community roads has been increasing due to the widespread use of ETC2.0. On the other hand, in terms of analysis methods for traffic accident prediction, multivariate analysis methods are expected to be applied to AI (Artificial Intelligence) for effective traffic accident analysis. Therefore, this study utilized ETC data and inverse reinforcement learning, a type of AI, to estimate the factors of potential hazards on community roads. We can obtaine knowledge on the possibility of using inverse reinforcement learning to estimate the effect of safety measures.