Using VR Generated Pedestrian Behaviors in Autonomous Vehicle Simulations

摘要

Simulation is a requirement for the safe and effective deployment of autonomous vehicle software given the wide range of possible scenarios which a vehicle can experience. These simulations need to accurately represent interactions with not just other vehicles, but also pedestrians. Humans often do not move in a perfectly linear way when interacting with the objects around them. Their position and rotations change as their focus and interest is piqued in various directions. Data collected by Deb et al. shows pedestrians in VR do not react the same to autonomous vehicles as they do to manned vehicles. Therefore, it is not suitable nor desirable to use straight line paths with pedestrian animations in simulation environments. In this experiment, we leverage the position and rotation data collected by Deb et al. during VR pedestrian crosswalk studies as a source for realistic pathing behaviors for animated pedestrians in the Mississippi Autonomous Vehicle Simulator.