Model-predictive control of the lane configuration at signalized intersections

Model-predictive control of the lane configuration at signalized intersections

Anema, F.J.C.

De Schutter, B. (mentor)
Van Katwijk, R.T. (mentor)

Mechanical, Maritime and Materials Engineering

DCSC

2015-09-22

The topic of this study is the control of signalized intersections, specifically the control of switching lanes, which are approach lanes of which the traffic assignment can be changed. When the demand at an intersection varies considerably, being able to change the lane configuration can reduce delay, especially in saturated conditions. Currently five intersections in the Netherlands already have a single switching lane that switches at two fixed times per day. However, using multiple switching lanes and being able to switch when needed may further reduce delay. Previous work on this topic was fairly limited in scope, did not use predictive control and did not integrate the control of traffic signals and switching lanes. In this study two model-predictive controllers for integrated traffic signal and switching lane control of a single intersection were designed. The first aims to minimize delay while the second aims to minimize the queue length. The main differences are that the first features a more elaborate prediction model and uses longer time step durations, while the second emphasizes faster measurements and computation and uses shorter time step durations. Both controllers were evaluated using PTV Vissim, a microscopic traffic simulation program. Fictional data consisting of simple signals as well as real-world data were used. In the fictional data tests, compared to the static situation, switching reduced delay for both controllers. In the real-world data tests no significant difference between static and switching was found. A state-of-the-art signal controller used for comparison performed equal or better in all cases. Considerable insight has been gained concerning the working of a dynamic lane configuration. The potential for delay reduction by using a predictive controller has been shown, but more research is needed to make the controllers suitable for handling real-world situations. The most important recommendation for future work is to improve the signal control part of both controllers.

control
intersection
traffic signals
dynamic lane configuration
model-predictive control

http://resolver.tudelft.nl/uuid:ac4174d1-ee21-43b1-a023-c4d99f28eebc

Student theses

master thesis

(c) 2015 Anema, F.J.C.