Contention-resolving model predictive control for an intelligent intersection traffic model

Abstract

We address the problem of optimally scheduling automated vehicles crossing an intelligent intersection by assigning vehicles with priorities and desired speed. An idealized intersection traffic model is established for the development and verification of the required algorithms. We formulate the intersection scheduling problem as a mixed integer programming (or MIP) problem which co-designs the priority and traveling speed for each vehicle. The co-design aims to minimize the vehicle waiting time at the intersection area, under a set of safety constraints. We derived a contention-resolving model predictive control (or MPC) algorithm to dynamically assign priorities and compute the vehicles’ traveling speeds. A branch cost formulation is proposed for the decision tree constructed by contention-resolving MPC based on time instants when collisions might occur among vehicles. Based on the priority assignments, a decentralized control law is designed to control each vehicle to travel with an optimal speed given a specific priority assignment. The optimal priority assignment can be determined by searching the lowest cost path in the decision tree. The solution computed by contention-resolving MPC is proved to be optimal given the condition of immediate access (or CIA) required in real-time scheduling. The effectiveness of the proposed method is verified through simulation and compared with the first-come-first-serve (or FCFS) and highest-speed-first (or HSF) scheduling strategies.