Towards network-wide safe and efficient traffic signal timing optimization based on costly stochastic simulation

Abstract

This study proposes a stochastic simulation-based network-wide signal timing optimization model with the balance consideration of traffic safety and efficiency, and solves it with a Bi-objective Stochastic Simulation-based Optimization (BOSSO) algorithm. During numerical experiments, an urban road network with 15 signalized and 5 non-signalized intersections in Changsha, China is modelled as an experimental scenario. The calibrated simulator VISSIM, Surrogate Safety Assessment Model (SSAM) and MATLAB are integrated to construct the VISSIM-SSAM-Matlab platform, based on which the network-wide signal timing optimization problems without and with coordination are solved by the BOSSO algorithm to trade-off traffic conflicts and total delay. Numerical results show that only hundreds of simulations are cost to obtain the ultimate non-dominated non-coordinated and coordinated signal timing plans, and validate a competing relationship between traffic conflicts and total delay. By the bi-objective comparison of three various signal plans from the overall and local aspects, the coordinated signal plan outperforms the non-coordinated one, which is followed by the field implemented one. It successfully demonstrates the effectiveness of the BOSSO method.