Intersection Navigation for Automated Vehicles

Abstract

This work provides a simulation environment for automated vehicles that drive through an intersection while using vehicle-to-vehicle (V2V) communications. To the end of demonstrating the proposed simulation environment, we provide software for risk assessment, which identifies dangerous situations by comparing what drivers intend to do with what they are expected to do in order to avoid collision. Our software takes into account uncertainties caused by vehicle state noise. Another way in which we demonstrate the proposed simulation environment is via software for coordinating intersection crossing in which vehicles request permissions from each other before deciding to cross, similar to the work of Casimiro and Schiller. The software is designed such that no two vehicles with conflicting maneuvers can be inside the intersection at the same time. Furthermore, we combined these two methods to create a system with multiple safety layers. The idea is that the risk assessment software works as a backup if some deviant driver does not adhere to the software for coordinating intersection crossing. The two methods are connected, such that whenever a vehicle is granted, the priority rules of the risk assessment software are temporarily changed in favor of the granted vehicle. Finally, we use the proposed evaluation environment for conducting experiments in a four-way give-way intersection. During these experiments, we observed that we can combine the risk assessment software with the coordination software. As we explain, the proposed simulation environment can help to measure to what degree collisions are avoided or mitigated, even while taking into account deviant vehicle behaviors, such as unpredictable traveling speeds, significant noise in state measurements, to name a few. We hope that the system can serve as a basis for future research projects.