Date Approved

6-11-2021

Embargo Period

6-11-2021

Document Type

Thesis

Degree Name

M.S. Computer Science

Department

Computer Science

College

College of Science & Mathematics

Advisor

Shen-Shyang Ho, PhD

Committee Member 1

Ning Wang, PhD

Committee Member 2

Anthony Breitzman, PhD

Keywords

Reinforcement Learning, Shared Mobility

Subject(s)

Machine learning; Bicycle sharing programs

Disciplines

Computer Engineering | Computer Sciences

Abstract

Shared mobility systems regularly suffer from an imbalance of vehicle supply within the system, leading to users being unable to receive service. If such imbalance problems are not mitigated some users will not be serviced. There is an increasing interest in the use of reinforcement learning (RL) techniques for improving the resource supply balance and service level of systems. The goal of these techniques is to produce an effective user incentivization policy scheme to encourage users of a shared mobility system to slightly alter their travel behavior in exchange for a small monetary incentive. These slight changes in user behavior are intended to over time increase the service level of the shared mobility system and improve user experience. In this thesis, two important questions are explored: (1) What state-action representation should be used to produce an effective user incentive scheme for a shared mobility system? (2) How effective are reinforcement learning-based solutions on the rebalancing problem under varying levels of resource supply, user demand, and budget? Our extensive empirical results based on data-driven simulation show that: 1. A state space with predicted user behavior coupled with a simple action mechanism produces an effective incentive scheme under varying environment scenarios. 2. The reinforcement learning-based incentive mechanisms perform at varying degrees of effectiveness under different environmental scenarios in terms of service level.

Share

COinS