AbstractCongestion on road traffic networks is a problem. Control theory can be applied to reduce congestion, and new technologies present more opportunities to do this. Strategies can be centralised or decentralised, but decentralised strategies offer advantages in terms of feasibility and scalability. We propose two decentralised control algorithms to be applied to road networks, controlling both vehicle routes and traffic lights. We validate these algorithms numerically using a microscopic traffic simulator.
We introduce the current literature on vehicle routing, and intersection control, providing an overview of each. We present a decentralised routing methodology, by which vehicles pick their routes by minimising a cost function based on travel time and road occupancy. We investigate the effect of tuning the control parameter which determines the relative balance between the two components of the cost function, and find that this is dependent on network topology and the presence of traffic lights. We find the algorithm performs favourably compared to the shortest path, and contrast our algorithm with Dynamic User Assignment using Gawron's iterative approach, showing the proposed method has distinct advantages in specific network topologies. We prove that in a scenario where all other vehicles are routed using only the shortest path, only a fraction of the vehicles in the network are required to adopt our proposed routing strategy to make a significant reduction in delays.
We present a modular decentralized traffic light controller and propose several algorithms which harness the potential of Vehicle-to-Infrastructure communication. We compare the performance of these algorithms with those already found in the literature, as well as exploring the impact of network topology on the performance of the controller. We find that our algorithms are able to outperform other controllers, but there is a significant relationship between network topology and algorithm performance. We then test these algorithms in the Luxembourg network and find the improvements in travel-time carry over to a real-world scenario.
|Date of Award||20 Mar 2018|
|Supervisor||Mario Di Bernardo (Supervisor)|