Modelling and control of urban junctions

Abstract

Continuous migration towards major cities around the world has brought an increase in the number of inhabitants in urban areas. With more people using the limited road infrastructure, mobility demands are exceeding the infrastructure capacity leading to increased traffic congestion. To address this problem a model predictive control scheme is developed and tested in this work to control in real time the traffic light timings. The scheme is based on a state space representation of the traffic dynamics through an urban traffic network. The proposed model incorporates a switching mechanism capturing both the nonlinear dynamics of normal traffic conditions and the linear evolution of vehicle queues during junction block-back scenarios. Moreover, while competing models suffer a quadratic increase in computational cost with every added junction, the proposed model exhibits only a linear increase in dimensionality and thus significantly lowers its computational demands. The simulation results from the proposed model are validated against a standard micro-modelling simulation package. A hierarchical control strategy based on this model is developed and tested in this dissertation to control in real time the traffic light timings of multiple urban junctions. The comparative advantages of the hierarchical controller over decentralised control are highlighted through an example. Results will show that with minimal computational power, communication requirements and infrastructure investment, the hierarchical control strategy manages to stabilize the queues on all the links in the network while improving the traffic flow through the network by minimizing the effect of junction block-back. The consistency of the results obtained are highlighted through Monte Carlo runs.