Elise Miller-Hooks
GMU


PhD Student:
Hossein Fotouhi

Hossein Fotouhi

Chief Data and Analytics Office, Advanced Analytics Center of Expertise, General Motors.

Email: hn.fotouhi@gmail.com

Dissertation Title:

“Crowdsourcing in Emerging Transportation Services: Models and Algorithms,” January 2021.

Ph.D. Dissertation Abstract:

Recent advances in technology and abundance of data have created opportunities for emergence of several new transportation services including, but not limited to, ride-hailing, parking, package, and meal delivery services. Many of these services use crowdsourced resources such as drivers as independent contractors. This dissertation addresses four problems of this type. These problems are directly related to crowdsourced transportation services and include a same-day delivery problem, a parking pricing problem, and two problems related to meal delivery services.

The same-day delivery time-guarantee (SDDTG) problem is proposed to assist online retailers who may not have a dedicated fleet of delivery vehicles, but can take advantage of crowdsourced delivery vehicles who are willing to perform delivery jobs for compensation. The SDDTG problem is formulated as a multi-stage, stochastic, mixed-integer program. The multi-stage stochastic program is approximated by a series of interconnected two-stage, stochastic programs, and a shrinking horizon framework is used to solve them sequentially as new information is revealed over time. To solve each of these two-stage stochastic programs, a scenario-based decomposition approach based on concepts of progressive hedging is proposed. The proposed model and algorithm is tested on a case study for Arlington County in Northern Virginia.

The optimal pricing in a mixed traditional and crowdsourced event parking market conceptualizes a parking market wherein personally owned crowdsourced parking spaces and other public parking locations compete to attract demand for an upcoming event. Uncertainty in both demand and supply is captured in the proposed stochastic framework. The proposed problem is formulated as a stochastic EPEC, where in the upper level parking owners set their time-differentiated parking prices to maximize their revenue. In response to these prices, in the lower level, users choose their parking locations seeking to minimize their total travel disutilities. A diagonalization technique with the gradient ascent method is proposed for its solution.

A third problem investigates the performance of meal delivery services with restrictions due to the ongoing COVID-19 pandemic. A scenario is considered where curbside pickup is regulated and the number of drivers simultaneously waiting at a restaurant to pick up orders is limited by local government to increase the safety of crowdsourced drivers and restaurant personnel. A few strategies, such as bundling orders and adding slack time to arrival and departure times for drivers to a restaurant reservation is proposed. The effectiveness of these strategies on the number and duration of curbside pickup violations as well as other metrics, such as the freshtime and click-to-door time of orders, is tested. Extensive numerical experiments were conducted using a discrete-event simulation conducted on a large real-world case study constructed using data provided by Grubhub, a leading meal delivery platform.

Crowdsourced drivers compete for jobs when entering the meal delivery and other similar services provided through the gig economy. The last problem takes a crowdsourced driver perspective, where a courier must determine, after each delivery job, a best location at which to preposition him/herself while waiting for the next job so as to maximize his/her opportunity to be offered future, profitable jobs. The daily operation of a meal delivery environment, including the arrival of orders and couriers, is simulated. The profit-maximization problem of a single courier is formulated as a Markov Decision Problem (MDP) and a machine-learning (Q-learning) algorithm is applied to determine the optimal policies (a function that returns the action to take for the driver’s location). The optimal policy obtained from Q-learning is compared to alternative policies, such as moving to the closest restaurant, that a driver might reasonably adopt.


M.S. Thesis Title (Completed 2016, University of Maryland):

Quantifying the Resilience of an Urban Traffic-electric Power Coupled System





Elise Miller-Hooks, Ph.D.
Professor
Bill & Eleanor Hazel Chair in Infrastructure Engineering

Phone: 703.993.1685
Email: miller@gmu.edu

Office: 4614 Nguyen Engineering Building

Address:
Sid and Reva Dewberry Department of Civil, Environmental and Infrastructure Engineering
George Mason University
4400 University Drive, MS 6C1
Fairfax, VA 22030
USA

 

Additional Resources

Students

Home

 


Volgenau School of Engineering
George Mason University
©