Safety and Liveness in Intelligent Intersections

Abstract

Automation of driving tasks is becoming of increasing interest for highway traffic management. Technologies for on-board sensing, combined with global positioning and inter-vehicular wireless communications, can potentially provide remarkable improvements in safety and efficiency. We address the problem of designing intelligent intersections where traffic lights and stop signs are removed, and cars negotiate the intersection through a combination of centralized and distributed decision making. Such intelligent intersections are representative of complex distributed hybrid systems which need architectures and algorithms with provable safety and liveness.

We propose a hybrid architecture which involves an appropriate interplay between centralized coordination and distributed freedom for the cars. Our approach is based on each car having an open-loop infinite horizon contingency plan, which is updated at each sampling time in a distributed fashion. We also define a partial order relation between cars which specifies to each car a set of cars whose worst case behaviors it should guard against. We prove the safety and liveness of the overall scheme. Concerning performance, we conduct a simulation study that shows the benefits over stop signs and traffic lights.

This material is based upon work partially supported by the Toyota Technical Center, Ann Arbor, MI, under Contract Number 2006-06246, NSF under Contract Nos. NSF ECCS-0701604, CNS-07-21992, NSF CNS 05-19535, and CCR-0325716, and USARO under Contract No. W-911-NF-0710287.