Scenario analyzer for real-time Dynamic Transportation Assignment (DTA) systems

• Main Author: Sui, Yihang
• Other Authors: Moshe E. Ben-Akiva.
• Format: Others
• Language: English
• Published: Massachusetts Institute of Technology 2018
• Subjects: Civil and Environmental Engineering.
• Online Access: http://hdl.handle.net/1721.1/119337

Thesis: S.M. in Transportation, Massachusetts Institute of Technology, Department of Civil and Environmental Engineering, 2018. === Cataloged from PDF version of thesis. === Includes bibliographical references (pages 71-74). === The optimization of network control strategies using real-time Dynamic Traffic Assignment systems typically utilizes short-term predictions of the network state within a rolling horizon framework. However, there exist several network control instruments (such as incentive schemes under daily budget constraints) whose optimization necessitate generating predictions beyond the "roll period" and for the entire day. This work addresses the aforementioned problem by proposing a "Scenario Analyzer" to extend the horizon for the optimization problem by providing relatively accurate predictions and forecasting results for the extended horizon. The Scenario Analyzer module adopts a data driven approach, and is formulated as a matching problem utilizing an archived historical database. The archived historical database includes the data from DTA systems as master data table, daily run table and historical scenario table. The matching algorithms use the historical scenario table and master data table to pair the current run feature(s) with historical runs feature(s); after finding a match, the current run will be stored at the daily run table. The matching problem may be solved using different statistical or machine learning algorithms, in terms of: 1) single time step feature matching 2) multiple time steps features matching. The performance of the proposed scenario analyzer is examined for the optimization of an app-based travel incentive scheme to reduce system wide energy consumption (referred to as Tripod) in the Boston CBD network. The k-NN and KL divergence matching algorithms are tested for a simulation period of 6 AM - 9 PM. Results indicate that the scenario analyzer with k-NN outperforms KLD algorithm probably because KLD need more data points to fully-develop the time-series properties. Among all the traffic features using in the matching algorithms, the cumulative energy consumption is the best indicator for similarity comparison. === by Yihang Sui. === S.M. in Transportation