Defensive or competitive Autonomous Vehicles: Which one interacts safely and efficiently with pedestrians?

The emergence of Autonomous Vehicles (AVs) could provoke unexpected challenges in urban traffic environments. One such crucial challenge is the conflicts between pedestrians and AVs, particularly on unsignalized mid-block crosswalks (UMC), where pedestrians are exposed to the AV flow. This study investigates the efficiency and safety performance of a UMC in the presence of both AVs and pedestrians considering the diversities in their behaviors. Through empirical analyses, two pedestrians’ crossing decision models are built and four groups of speed profiles are classified. Meanwhile, based on previous literature, defensive and competitive driving strategies are assumed for AVs. The simulation is implemented on an agent-based framework that can dynamically reproduce the kinematic interactions between pedestrians and vehicles. Results indicated that with a reasonable safety margin (2.5 s), percentages of low post encroachment time events for competitive AVs with different pedestrian types are smaller than defensive AVs with differences of 0.2% to 2.9%. The average delays of competitive AVs for all pedestrian types are smaller than defensive AVs with a maximum estimated difference of 39 s. Moreover, the analysis showed that lowering the speed limit may reduce the crash rate of competitive AV up to 0%. It is also found that the pedestrians who make reckless crossing decisions and change their speed drastically during the crossing process are more likely to incur crashes with competitive AVs. Therefore, if pedestrian behaviors can be regulated reasonably, competitive AVs with appropriate parameter settings are most suitable for UMC in the future.