Date on Master's Thesis/Doctoral Dissertation


Document Type

Doctoral Dissertation

Degree Name

Ph. D.


Industrial Engineering

Degree Program

Industrial Engineering, PhD

Committee Chair

Bai, Lihui

Committee Co-Chair (if applicable)

Alexander, Suraj

Committee Member

Alexander, Suraj

Committee Member

Bae, Kihwan

Committee Member

Guan, Jian

Author's Keywords

crowdsourcing transportation; many-to-many assignment problem; data mining; predictive modeling


Collaborative transportation platforms have emerged as an innovative way for firms and individuals to meet their transportation needs through using services from external profit-seeking drivers. A number of collaborative transportation platforms (such as Uber, Lyft, and MyDHL) arise to facilitate such delivery requests in recent years. A particular collaborative transportation platform usually provides a two sided marketplace with one set of members (service seekers or passengers) posting tasks, and the another set of members (service providers or drivers) accepting on these tasks and providing services. As the collaborative transportation platform attracts more service seekers and providers, the number of open requests at any given time can be large. On the other hand, service providers or drivers often evaluate the first couple of pending requests in deciding which request to participate in. This kind of behavior made by the driver may have potential detrimental implications for all parties involved. First, the drivers typically end up participating in those requests that require longer driving distance for higher profit. Second, the passengers tend to overpay under a competition free environment compared to the situation where the drivers are competing with each other. Lastly, when the drivers and passengers are not satisfied with their outcomes, they may leave the platforms. Therefore the platform could lose revenues in the short term and market share in the long term. In order to address these concerns, a decision-making support procedure is needed to: (i) provide recommendations for drivers to identify the most preferable requests, (ii) offer reasonable rates to passengers without hurting driver’s profit. This dissertation proposes a mathematical modeling approach to address two aspects of the crowdsourcing ridesharing platform. One is of interest to the centralized platform management on the assignment of requests to drivers; and this is done through a multi-criterion many to many assignment optimization. The other is of interest to the decentralized individual drivers on making optimal bid for multiple assigned requests; and this is done through the use of prospect theory. To further validate our proposed collaborative transportation framework, we analyze the taxi yellow cab data collected from New York city in 2017 in both demand and supply perspective. We attempt to examine and understand the collected data to predict Uber-like ridesharing trip demands and driver supplies in order to use these information to the subsequent multi-criterion driver-to-passenger assignment model and driver's prospect maximization model. Particularly regression and time series techniques are used to develop the forecasting models so that centralized module in the platform can predict the ridesharing demands and supply within certain census tracts at a given hour. There are several future research directions along the research stream in this dissertation. First, one could investigate to extend the models to the emerging concept of "Physical Internet" on commodity and goods transportation under the interconnected crowdsourcing platform. In other words, integrate crowdsourcing in prevalent supply chain logistics and transportation. Second, it's interesting to study the effect of Uber-like crowdsourcing transportation platforms on existing traffic flows at the various levels (e.g., urban and regional).