Please use this identifier to cite or link to this item: http://scholarbank.nus.edu.sg/handle/10635/130183
Title: HIGH DIMENSIONAL BAYESIAN OPTIMIZATION WITH APPLICATION TO TRAFFIC SIMULATION
Authors: SON JAEMIN
Keywords: Toll Optimization in Traffic Simulation, Bayesian Optimization
Issue Date: 13-Jul-2016
Citation: SON JAEMIN (2016-07-13). HIGH DIMENSIONAL BAYESIAN OPTIMIZATION WITH APPLICATION TO TRAFFIC SIMULATION. ScholarBank@NUS Repository.
Abstract: In Singapore, ERP (Electronic Road Pricing) gantries are installed on selected road segments in an attempt to control traffic congestion through monetary incentives. Decision makers are interested in improving drivers' utility and reducing travel time through optimization of ERP rates in simulation. These two objectives are likely non-linear and non-convex as they are the aggregate effect of traffic flow in individual road segments. Thus, gradient descent, linear programming and convex optimization do not apply straightforwardly. Also, an urban scale traffic simulation incurs high computational costs that fewer function evaluations are preferred during the optimization process. Several approaches have been proposed including heuristic search methods and target-based optimization, however, they lack theoretical guarantee or ignore coordination between different gantries. We introduce Bayesian Optimization (BO) to deal with such challenges. BO constructs surrogates of the objective function and evaluates the most promising points based on the surrogates. With flexible surrogates such as Gaussian Processes (GP), BO can optimize black-box functions that are non-linear and non-convex and reduce the number of function evaluations effectively without compromising optimization performance. Also, we address challenges with high dimensionality of toll optimization in urban scale. In fact, the conventional BO has only been successful in input dimension less than 10. In order to deal with input dimension up to a hundred, we use existing high dimensional BOs that are introduced in machine learning community and new model that captures hidden coordination between gantries.
URI: http://scholarbank.nus.edu.sg/handle/10635/130183
Appears in Collections:Master's Theses (Open)

Show full item record
Files in This Item:
File Description SizeFormatAccess SettingsVersion 
SonJM.pdf3.59 MBAdobe PDF

OPEN

NoneView/Download

Page view(s)

77
checked on Feb 8, 2019

Download(s)

1
checked on Feb 8, 2019

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.