Real-time Personalized Tolling with Long-term Objectives
Managed lanes are separate tolled lanes adjacent to free general-purpose lanes. The key real-time operation problem is how to set the toll for both effective network management and revenue generation, jointly considering the objectives of the operator, the travelers and the regulator. Based on a com...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Published: |
Massachusetts Institute of Technology
2022
|
| Online Access: | https://hdl.handle.net/1721.1/143404 |
| _version_ | 1826188352558202880 |
|---|---|
| author | Xie, Yifei |
| author2 | Moshe Ben-Akiva |
| author_facet | Moshe Ben-Akiva Xie, Yifei |
| author_sort | Xie, Yifei |
| collection | MIT |
| description | Managed lanes are separate tolled lanes adjacent to free general-purpose lanes. The key real-time operation problem is how to set the toll for both effective network management and revenue generation, jointly considering the objectives of the operator, the travelers and the regulator. Based on a comprehensive analysis of travel behavior, this thesis develops a solution with adaptive personalized pricing.
Travelers are observed to either predominantly use managed lanes or almost never. This could be attributed to two competing latent behavioral factors: preference heterogeneity, and state dependence—not switching between options causally yields positive utility. Their econometric quantifications have crucial implications on pricing, but are challenging due to endogeneity known as the initial condition problem. We begin by proposing a Control Function solution under a general setting, which is shown to improve a commonly used solution by Wooldridge. Then, through applying the developed solutions to empirical data, we discovered heterogeneity and state dependence to be both significant in explaining the usage decision. It is further shown that when ignoring unobserved heterogeneity or the initial condition problem, state dependence will be largely overstated. Price endogeneity caused by dynamic pricing is also discovered and corrected.
The developed behavioral model is integrated into an online personalized tolling system that incorporates prediction, optimization and personalization. In addition to optimizing the toll adaptively, an online bi-level optimization problem is formulated to jointly offer personalized discounts. A flexible multi-component objective is designed to consider not only short-term revenue and social welfare, but also the impact on future revenue based on the state-dependent choice behavior. The online personalized tolling system is deployed to a microscopic traffic simulator calibrated with real data. The results show simultaneous improvements of revenue, traffic conditions and social welfare. Equity improvement is also discovered as travelers with lower values of time are presented lower tolls.
The developed methodologies for behavioral analysis and personalized pricing could be directly adapted for other applications in transportation and beyond. |
| first_indexed | 2024-09-23T07:58:04Z |
| format | Thesis |
| id | mit-1721.1/143404 |
| institution | Massachusetts Institute of Technology |
| last_indexed | 2024-09-23T07:58:04Z |
| publishDate | 2022 |
| publisher | Massachusetts Institute of Technology |
| record_format | dspace |
| spelling | mit-1721.1/1434042022-06-16T04:00:07Z Real-time Personalized Tolling with Long-term Objectives Xie, Yifei Moshe Ben-Akiva Ravi Seshadri Massachusetts Institute of Technology. Department of Civil and Environmental Engineering Managed lanes are separate tolled lanes adjacent to free general-purpose lanes. The key real-time operation problem is how to set the toll for both effective network management and revenue generation, jointly considering the objectives of the operator, the travelers and the regulator. Based on a comprehensive analysis of travel behavior, this thesis develops a solution with adaptive personalized pricing. Travelers are observed to either predominantly use managed lanes or almost never. This could be attributed to two competing latent behavioral factors: preference heterogeneity, and state dependence—not switching between options causally yields positive utility. Their econometric quantifications have crucial implications on pricing, but are challenging due to endogeneity known as the initial condition problem. We begin by proposing a Control Function solution under a general setting, which is shown to improve a commonly used solution by Wooldridge. Then, through applying the developed solutions to empirical data, we discovered heterogeneity and state dependence to be both significant in explaining the usage decision. It is further shown that when ignoring unobserved heterogeneity or the initial condition problem, state dependence will be largely overstated. Price endogeneity caused by dynamic pricing is also discovered and corrected. The developed behavioral model is integrated into an online personalized tolling system that incorporates prediction, optimization and personalization. In addition to optimizing the toll adaptively, an online bi-level optimization problem is formulated to jointly offer personalized discounts. A flexible multi-component objective is designed to consider not only short-term revenue and social welfare, but also the impact on future revenue based on the state-dependent choice behavior. The online personalized tolling system is deployed to a microscopic traffic simulator calibrated with real data. The results show simultaneous improvements of revenue, traffic conditions and social welfare. Equity improvement is also discovered as travelers with lower values of time are presented lower tolls. The developed methodologies for behavioral analysis and personalized pricing could be directly adapted for other applications in transportation and beyond. Ph.D. 2022-06-15T13:18:28Z 2022-06-15T13:18:28Z 2022-02 2022-04-29T19:17:22.411Z Thesis https://hdl.handle.net/1721.1/143404 In Copyright - Educational Use Permitted Copyright MIT http://rightsstatements.org/page/InC-EDU/1.0/ application/pdf Massachusetts Institute of Technology |
| spellingShingle | Xie, Yifei Real-time Personalized Tolling with Long-term Objectives |
| title | Real-time Personalized Tolling with Long-term Objectives |
| title_full | Real-time Personalized Tolling with Long-term Objectives |
| title_fullStr | Real-time Personalized Tolling with Long-term Objectives |
| title_full_unstemmed | Real-time Personalized Tolling with Long-term Objectives |
| title_short | Real-time Personalized Tolling with Long-term Objectives |
| title_sort | real time personalized tolling with long term objectives |
| url | https://hdl.handle.net/1721.1/143404 |
| work_keys_str_mv | AT xieyifei realtimepersonalizedtollingwithlongtermobjectives |