Investigating the Efficiency and Safety of Signalized Intersections Under Mixed Flow Conditions of Autonomous and Human-Driven Vehicles

Abstract

© 2020, King Fahd University of Petroleum & Minerals. Field experiments of autonomous driving systems have already been started in several countries around the world. However, investigations on the performance of signalized intersections under the mixed flow condition of autonomous vehicle (AV) and human-driven vehicle (HDV) are still very limited, especially in terms of considering the interaction between vehicles. This study aims at developing a methodology to evaluate the efficiency and safety performance of signalized intersections under the conditions of AV mixed flows. If AVs are set aggressively such as small critical gap acceptance thresholds, approach capacity of signalized intersections may increase. However, the aggressive driving maneuver and the misjudgment by human drivers to AV may lead to some potential safety risks. On the other hand, if AVs are set conservatively to ensure large safety margins, for example by setting large critical gap thresholds or a longer following headway, a significant drop in the capacity is expected while the safety improvements may not be insured since extra delays will be imposed on HDVs which may induce risky behavior. This study finds that AVs with either excessively aggressive or conservative settings will lead to waste of capacity or potential safety hazard. AVs with aggressive settings will enable them to accept a short gap or lag, while AVs with conservative settings will increase the waiting time of the following HDVs, which may induce aggressive driving behavior. Both types of AV settings may contribute to safety risks represented by short time-to-collision and short post encroachment time.