The Optimisation of Traffic Signal Control Over a Rolling Horizon

Abstract

Discrete time forms of adaptive traffic signal control which optimise over a rolling horizon are focused on. Markov chain techniques to compute the evolution of expected queues require prohibitively large amounts of memory and processing. Expressions are developed for the propagation forward in time of the first and second moments of queue size. A bulk service queuing model with random arrivals and departures is used. Particular forms of distribution are not required, but binomial arrivals and departures were used in simulation. Initial results suggest that a short rolling horizon is adequate in the absence of public transport priority.