A Fairness Oriented Reinforcement Learning Approach for the Operation and Control of Shared Micromobility Services

This thesis explores Reinforcement Learning (RL) approaches to enhance fairness in Shared Micromobility Services, with a case study based on bike-sharing systems. Traditional sharing models often grapple with the conflict between maximising profitability and providing equitable service coverage. This research aims to address this challenge by integrating RL to optimise the distribution and availability of bikes, ensuring accessibility across diverse urban areas, including underserved communities. The core of the thesis involves developing and testing a model that can adjust bike distribution in response to data on usage patterns, demands and geography. By leveraging RL techniques, the study not only predicts high-demand areas but also proactively manages resource allocation to balance profit and fair coverage. Empirical evaluations are conducted using a simulated environment based on real-world urban layouts and usage data. These experiments demonstrate the efficacy of RL in balancing the tradeoff between service fairness and profitability. The thesis contributes to the growing body of knowledge in applying RL aimed at fairness, particularly in urban planning and sustainable transportation. It offers a perspective on longstanding issues for more responsible urban mobility solutions.