Self-tuning model predictive control for signalized traffic junctions

Traffic signal timing plays an important role in ensuring efficient flow and reduction of traffic congestion. Fixed signal times work well when traffic conditions are consistent. However performance degrades when traffic conditions are subject to high demands or during unusual occurrences such as traffic incidents or unanticipated network obstructions, causing significant changes to the normal traffic conditions. To address these situations, several traffic-responsive systems were proposed. However in these cases, the controller parameters are typically set at installation and not adaptively tuned to changing traffic behaviour. Hence, in this work, a novel adaptive control system which can self-tune to respond to changing traffic conditions is developed, leading towards an autonomic system. This system makes use of a MPC based approach which can autonomously tune the controller to ensure good performance despite changing traffic conditions. Different norms were tested as objective functions for the optimization problem. Results highlight the effectiveness of the proposed traffic light timing controller.