Autonomous Intersection Management systems: criteria, implementation and evaluation

Abstract

The foreseeable large-scale deployment of autonomous vehicles in the near future raises the question of autonomous intersection management (AIM). Numerous AIM designs have been proposed, but they lack a common vision of what defines a good system. In this study, the authors discuss a set of conflicting evaluation criteria that need to be balanced in the design of an AIM system, but are often considered individually in the literature. They then introduce their own priority-based design, where an intersection controller assigns priorities to incoming vehicles. On being assigned a priority, vehicles then cross the intersection while maintaining a so-called brake-safe state with respect to higher priority vehicles, rendering the system robust. They have performed extensive simulations to showcase the properties of the proposed system, and notably that it satisfactorily balances their criteria while remaining efficient.