Study on situation awareness for autonomous driving safety in the mixed traffic

Abstract

The greatest barriers for the realization of the vehicle autonomy are the uncertainties that affect the safety of the autonomous driving. Especially, the uncertainty at the intersection is amplified in proportion to the complexity of the layout. And the crossed and interrupted traffic also contribute to the uncertainty at the intersection. Moreover, in the transition period, the autonomous vehicles are mixed with the non-autonomous vehicles which are the great uncertainty by themselves. Therefore, it is almost impossible for the autonomous vehicle to resolve the problems alone, and it requires the assist of the infrastructure such as the Cooperative Intelligent Transportation System (C-ITS). The edge server of C-ITS has more computing power compared to the autonomous vehicle, and especially C-ITS can be equipped with the sensor network of omniscient view that can monitor the mixed traffic at the intersection without any obstruction. The situation awareness is to unveil the uncertainties, and it can be done by both of infrastructure side and autonomous vehicle side. Among the various uncertainties in the autonomous driving, the trajectory of surrounding vehicle and road friction are the most significant ones that are relevant to the safety. From this background, this study introduces first, the situation awareness on traffic, which is to predict the trajectory of the non-autonomous vehicle in the mixed traffic of intersection at the infrastructure side, and second, the situation awareness on road, which is to estimate the road friction at the autonomous vehicle side. The former utilizes the multiple hypotheses on the maneuver set, and the proposed all-in-one framework is possible to make a stable trajectory prediction by virtue of the interaction between them. The latter proposes an on-line road friction estimation method which is integrated as a part of the Autonomous Emergency Braking (AEB) system. The proposed method is possible to estimate the road friction which is valid at the instance of emergency operation in on-line, which is not introduced in the literatures yet. The proposed frameworks are evaluated with the experiments. Additionally, the possibility of cooperative situation awareness between the infrastructure and autonomous vehicle is introduced in Appendix.