Abstract
Traffic accidents on community roads with occlusions are one of the unsolved social issues in Japan because advanced driver assistance systems and autonomous driving systems only using onboard sensors cannot behave adequately due to the occlusions caused by house walls. Although connected safety system is promising to prevent traffic accidents in such situations, placing roadside sensors at every intersection costs too much. Motivated by above backgrounds, this study proposes a connected collision avoidance system via a stochastic localization method that combines observation information from roadside sensors and onboard/portable devices of traffic participants for realizing the safety systems as well as reducing the necessary number of roadside sensors. To evaluate the proposed system from the viewpoints of uncertainty of the observation due to the layout and the specification of roadside sensors, this study conducts simulations in the course that modeled community roads in the actual town. As a result, this study confirms that the proposed method realizes the collision avoidance between an automobile and a bicycle even on a condition of a sparse layout of the roadside sensors owing to the concept of the stochastic localization based on the extended Kalman filter. The findings of this study will contribute to realization of the connected safety system on community roads with occlusions under the reasonable implementation conditions that roadside sensors will not be fully installed at every intersection.
