Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Small Fleet, Big Impact: Enhancing Shared Micromobility Efficiency through Minimal Autonomous Vehicle Deployment

Heng Tan, Hua Yan, Lucas Yang, Yu Yang

TL;DR

This paper tackles spatio-temporal demand imbalance in shared micromobility by introducing SMART, a two-level hierarchical reinforcement learning framework that minimally deploys autonomous shared micromobility vehicles to complement existing scheduling. The approach couples a high-level ASMV initial deployment module with a low-level self-rebalancing policy, both trained via PPO with alternating updates on real-world Chicago data. Results show that deploying around 3% ASMVs yields substantial gains in demand satisfaction across multiple baselines and conditions while remaining robust to faults and background demand, and it preserves the integrity of existing scheduling strategies. The work demonstrates a practical, cost-conscious pathway to more responsive urban micromobility systems and lays groundwork for broader integration of ASMVs into city-scale transportation planning.

Abstract

Shared micromobility systems, such as electric scooters and bikes, have gained widespread popularity as sustainable alternatives to traditional transportation modes. However, these systems face persistent challenges due to spatio-temporal demand fluctuations, often resulting in a mismatch between vehicle supply and user demand. Existing shared micromobility vehicle scheduling methods typically redistribute vehicles once or twice per day, which makes them vulnerable to performance degradation under atypical conditions. In this work, we design to augment existing micromobility scheduling methods by integrating a small number of autonomous shared micromobility vehicles (ASMVs), which possess self-rebalancing capabilities to dynamically adapt to real-time demand. Specifically, we introduce SMART, a hierarchical reinforcement learning framework that jointly optimizes high-level initial deployment and low-level real-time rebalancing for ASMVs. We evaluate our framework based on real-world e-scooter usage data from Chicago. Our experiment results show that our framework is highly effective and possesses strong generalization capability, allowing it to seamlessly integrate with existing vehicle scheduling methods and significantly enhance overall micromobility service performance.

Small Fleet, Big Impact: Enhancing Shared Micromobility Efficiency through Minimal Autonomous Vehicle Deployment

TL;DR

This paper tackles spatio-temporal demand imbalance in shared micromobility by introducing SMART, a two-level hierarchical reinforcement learning framework that minimally deploys autonomous shared micromobility vehicles to complement existing scheduling. The approach couples a high-level ASMV initial deployment module with a low-level self-rebalancing policy, both trained via PPO with alternating updates on real-world Chicago data. Results show that deploying around 3% ASMVs yields substantial gains in demand satisfaction across multiple baselines and conditions while remaining robust to faults and background demand, and it preserves the integrity of existing scheduling strategies. The work demonstrates a practical, cost-conscious pathway to more responsive urban micromobility systems and lays groundwork for broader integration of ASMVs into city-scale transportation planning.

Abstract

Shared micromobility systems, such as electric scooters and bikes, have gained widespread popularity as sustainable alternatives to traditional transportation modes. However, these systems face persistent challenges due to spatio-temporal demand fluctuations, often resulting in a mismatch between vehicle supply and user demand. Existing shared micromobility vehicle scheduling methods typically redistribute vehicles once or twice per day, which makes them vulnerable to performance degradation under atypical conditions. In this work, we design to augment existing micromobility scheduling methods by integrating a small number of autonomous shared micromobility vehicles (ASMVs), which possess self-rebalancing capabilities to dynamically adapt to real-time demand. Specifically, we introduce SMART, a hierarchical reinforcement learning framework that jointly optimizes high-level initial deployment and low-level real-time rebalancing for ASMVs. We evaluate our framework based on real-world e-scooter usage data from Chicago. Our experiment results show that our framework is highly effective and possesses strong generalization capability, allowing it to seamlessly integrate with existing vehicle scheduling methods and significantly enhance overall micromobility service performance.

Paper Structure

This paper contains 27 sections, 10 equations, 11 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Preliminary and Motivation
  3. Data Description
  4. Why Autonomous Vehicle Scheduling?
  5. Problem Formulation
  6. Methodology
  7. Design Overview
  8. Two-level Hierarchical RL-based vehicle scheduling
  9. Training Method
  10. Evaluation
  11. Experiment Setup
  12. Implementation
  13. Baselines
  14. Metric
  15. Overall Performance
  16. ...and 12 more sections

Figures (11)

  • Figure 1: The average demand satisfaction rate of different baselines in two months
  • Figure 2: The differences in trip number between low-performance and other days from the spatial perspective
  • Figure 3: The differences in trip number between low-performance days and other days from the temporal perspective
  • Figure 4: The distribution of system performance improvements between low-performance days and other days
  • Figure 5: Overview of hierarchical RL framework for rebalancing autonomous shared micromobility vehicles
  • ...and 6 more figures