Integrated Optimization of Transportation Demand Management and Traffic Operations Using Bootstrapped Support Vector Regression
Integrating optimization of transportation demand management and traffic operations strategies has been challenging since analytical models that are capable of predicting the joint impact of these strategies can hardly be formulated. On the other hand, simulation tools can capture the performance of transportation systems in response to combinations of various policies at different levels, while they are more often used for evaluation instead of optimization due to their high computational cost. This paper applies the surrogate-based optimization method, which enjoys the advantages of both simulation evaluations and mathematical optimization methods, to jointly optimize a transportation demand management policy (i.e. high-occupancy/toll, HOT) and a traffic operational strategy (i.e. freeway diversion control) for a freeway-arterial corridor (I-270 corridor) under a work zone scenario in the State of Maryland. A distribution based support vector regression model (D-SVR) is developed to take into account the observed asymmetric distribution of simulation noise. In addition, bootstrapping and expected improvement based infill strategies are incorporated into the D-SVR for the global optimization. Results of the numerical test show that the D-SVR can approximate the true response surface more accurately and the bootstrapped D-SVR converges to the global optimal solution much faster than other models. By applying the optimal solution predicted by the D-SVR in the I-270 corridor, the average trip travel time for work zone impacted vehicles is reduced by 7.66% compared to baseline during the PM peak period. With negligible cost in infrastructure construction, the total saved travel time by the D-SVR predicted optima corresponds to a yearly saving of around $11 million for the 5-hour extended PM peak period.
-
Supplemental Notes:
- This paper was sponsored by TRB committee ADB30 Transportation Network Modeling.
-
Corporate Authors:
500 Fifth Street, NW
Washington, DC United States 20001 -
Authors:
- He, Xiang
- Chen, Xiqun (Michael)
- Xiong, Chenfeng
- Zhu, Zheng
- Zhang, Lei
-
Conference:
- Transportation Research Board 94th Annual Meeting
- Location: Washington DC, United States
- Date: 2015-1-11 to 2015-1-15
- Date: 2015
Language
- English
Media Info
- Media Type: Digital/other
- Features: Figures; References; Tables;
- Pagination: 22p
- Monograph Title: TRB 94th Annual Meeting Compendium of Papers
Subject/Index Terms
- TRT Terms: High occupancy toll lanes; Highway operations; Optimization; Peak periods; Regression analysis; Traffic diversion; Traffic simulation; Travel demand management; Travel time; Variable message signs; Work zones
- Geographic Terms: Maryland
- Subject Areas: Highways; Operations and Traffic Management; Planning and Forecasting; I72: Traffic and Transport Planning; I73: Traffic Control;
Filing Info
- Accession Number: 01558275
- Record Type: Publication
- Report/Paper Numbers: 15-5500
- Files: TRIS, TRB, ATRI
- Created Date: Mar 30 2015 9:35AM