Project Description

Congestion has become a global phenomenon, in particular in great urban areas in which daily traffic jams are in most cases a major concern. Managing signal plans efficiently is one of the most cost-effective methods. However, existing signal control strategies are less powerful in handling congested network with spillbacks and grid-type topology.

Enhancing the reliability of our networks is currently recognized as a critical goal in the US and in Europe. There is extensive evidence that indicates that travel time reliability is accounted by travelers in a variety of travel decisions, such as departure time and route choice. Hence, operating our networks such as to reduce both the average and the variability of trip travel times would be highly valued by travelers. However, urban traffic management strategies are typically formulated such as to improve first-order performance metrics (e.g. expected trip travel times, expected link speeds). The main challenge in addressing reliability in traditional transportation optimization problems is the need to provide an accurate analytical and tractable approximation of trip travel time distribution, or of its first- and second-order moments.

The complex between-link spatial-temporal dependency patterns makes accurate analytical modeling of urban road networks a challenge. In particular, when the aim is to model metrics related to the paths chosen by the drivers, in order to reflect driver experiences. Thus, this work proposes new signal control strategies for large-scale congested urban networks that can tackle these challenges.

In this thesis, a simulation-based optimization (SO) is used to address traffic signal control problems. Microscopic simulators describe in detail the interactions between vehicle performance, traveler behavior and the underlying transportation infrastructure. They can ultimately contribute to the design of traffic management strategies, providing detailed system performance estimates to infer the design and operations of urban networks. To ensure the computational efficiency, an analytical approximation of objective function is needed. We develop different formulations of travel time reliability based on both link travel time and path or trip travel time distributional information, and then use those formulations in signal design strategies to fulfill the reliability requirements. We also design a simulation-based adaptive traffic signal control algorithm to adjust signals plans dynamically according to real-time traffic conditions.

We apply the reliable signal control strategy to both city center and the full city of Lausanne. The proposed simulation-based adaptive traffic signal control algorithm is applied to a grid-type urban network with heavy traffic in east Manhattan area (New York City, USA). In both cases, proposed methods lead to signal plan with better performance in terms of various performance metrics.

Research Team

• Merkebe Demissie
• Gonçalo Correia (supervisor)

Financial Support

• FCT

Stage of Progress

• Finished in 2014