Optimal traffic signal control system using an artificial neural network.
Item
-
Title
-
Optimal traffic signal control system using an artificial neural network.
-
Identifier
-
AAI9908313
-
identifier
-
9908313
-
Creator
-
Fan, Jianzhong.
-
Contributor
-
Adviser: Mitsuru Saito
-
Date
-
1998
-
Language
-
English
-
Publisher
-
City University of New York.
-
Subject
-
Engineering, Civil | Artificial Intelligence | Urban and Regional Planning
-
Abstract
-
This dissertation research aims to develop an artificial intelligence based signal timing system. The system was intended to dynamically control a traffic signal by finding an optimal signal timing to minimize delay at signalized intersections within a few seconds. This optimal traffic signal control system (OTSCS) consists of two major components: an artificial neural network designed for analyzing level of service at a signalized intersection and a heuristic search.;Level of service analysis neural network (LOSANN) is a flat connection network with an error back-propagation algorithm. In order to sufficiently represent the traffic environment, LOSANN has 132 neurons at its input layer. Such a large number of input neurons tend to make the learning process relatively difficult and time consuming. Therefore, a variety of network architectures, learning modes and learning rate types are studied in this dissertation. Two learning rate types, linear and weighted exponential learning rate, are created to enhance the learning ability of LOSANN. The output of LOSANN is the average stopped delay per vehicle at an intersection.;An optimum traffic signal timing model (OTSTM) integrates LOSANN to obtain a signal timing which minimizes the average stopped delay at an intersection. A new heuristic search strategy named "direction search" is created in this study. It has an order of growth of O(n{dollar}\sp2{dollar}) in the worst case. The n is the number of traffic signal cycle lengths that could be used at the specific intersection. The numeral examples tested in the research show that the search speed of direction search is more the 10 times faster than that of the conventional depth-first search and the solutions of these two searches are very close.;OTCSC is programmed by Visual Basic Release 5.0 as comprehensive research tool. It provides options for selecting artificial neural network models and optimizing traffic signal models. OTSCS can be used to analyze level of service at an intersection, optimize signal timing, and understand the learning behavior of neural networks and the features of heuristic searches.
-
Type
-
dissertation
-
Source
-
PQT Legacy CUNY.xlsx
-
degree
-
Ph.D.
