Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

Matteo Cederle, Luca Vittorio Piron, Marina Ceccon, Federico Chiariotti, Alessandro Fabris, Marco Fabris, Gian Antonio Susto

TL;DR

The paper tackles spatial fairness in dockless shared micromobility by formulating a factorized multi-agent MDP and applying Q-learning to optimize rebalancing. It introduces a fairness-aware reward that penalizes inequitable service across city-area categories, quantified with a Gini index, and demonstrates Pareto-front trade-offs between cost and fairness using synthetic case studies. The key contribution is a scalable, separable RL framework that directly controls equity via a beta-weighted fairness term, showing that meaningful fairness gains can be achieved with moderate cost increases. The work lays groundwork for future extensions to time-varying demand and inter-area dependencies, and provides publicly available code for replication.

Abstract

As Machine Learning grows in popularity across various fields, equity has become a key focus for the AI community. However, fairness-oriented approaches are still underexplored in smart mobility. Addressing this gap, our study investigates the balance between performance optimization and algorithmic fairness in shared micromobility services providing a novel framework based on Reinforcement Learning. Exploiting Q-learning, the proposed methodology achieves equitable outcomes in terms of the Gini index across different areas characterized by their distance from central hubs. Through vehicle rebalancing, the provided scheme maximizes operator performance while ensuring fairness principles for users, reducing iniquity by up to 85% while only increasing costs by 30% (w.r.t. applying no equity adjustment). A case study with synthetic data validates our insights and highlights the importance of fairness in urban micromobility (source code: https://github.com/mcederle99/FairMSS.git).

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

TL;DR

The paper tackles spatial fairness in dockless shared micromobility by formulating a factorized multi-agent MDP and applying Q-learning to optimize rebalancing. It introduces a fairness-aware reward that penalizes inequitable service across city-area categories, quantified with a Gini index, and demonstrates Pareto-front trade-offs between cost and fairness using synthetic case studies. The key contribution is a scalable, separable RL framework that directly controls equity via a beta-weighted fairness term, showing that meaningful fairness gains can be achieved with moderate cost increases. The work lays groundwork for future extensions to time-varying demand and inter-area dependencies, and provides publicly available code for replication.

Abstract

As Machine Learning grows in popularity across various fields, equity has become a key focus for the AI community. However, fairness-oriented approaches are still underexplored in smart mobility. Addressing this gap, our study investigates the balance between performance optimization and algorithmic fairness in shared micromobility services providing a novel framework based on Reinforcement Learning. Exploiting Q-learning, the proposed methodology achieves equitable outcomes in terms of the Gini index across different areas characterized by their distance from central hubs. Through vehicle rebalancing, the provided scheme maximizes operator performance while ensuring fairness principles for users, reducing iniquity by up to 85% while only increasing costs by 30% (w.r.t. applying no equity adjustment). A case study with synthetic data validates our insights and highlights the importance of fairness in urban micromobility (source code: https://github.com/mcederle99/FairMSS.git).
Paper Structure (11 sections, 2 theorems, 12 equations, 3 figures, 1 table)

This paper contains 11 sections, 2 theorems, 12 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. System model
  3. MDP formulation
  4. Factorized MDP representation
  5. Fairness-oriented design
  6. Fairness metrics in MSS
  7. Fair Reinforcement Learning solution
  8. Numerical simulations
  9. Pareto fronts for the proposed approach
  10. Fairness and costs trends for the five-category scenario
  11. Conclusions and future directions

Key Result

Proposition 1

Given the global reward function in eq:global_reward, the optimal solution to the multi-agent MDP defined by Problem prob:multi is given by the combination of the individual solutions to the agent problems in Problem prob:single. The resulting solution then enjoys the convergence properties of singl

Figures (3)

  • Figure 1: Pareto fronts for the considered bi-objective optimization problem. The cost minimization and the fairness maximization objectives are represented on the x and y axes, respectively. Each mark corresponds to a different value of $\beta$, and the Pareto front (in blue) includes all efficient solutions. The red points correspond to Pareto-inefficient solutions.
  • Figure 2: Convergence behavior of the algorithm over training (mean $\pm 1.96$ standard dev.) for representative values of $\beta$.
  • Figure 3: Distributions over 10 Monte Carlo runs of fairness and cost-related performance metrics as a function of $\beta$. (d): The yellow dashed line and area indicate resp. average and $97.5$th percentile of the distribution of the initial number of vehicles.

Theorems & Definitions (4)

  • Proposition 1: Separability of the MSS problem becker2004solving
  • proof
  • Proposition 2
  • proof