Scholarlink

<p>Shared Autonomous Mobility Services (SAMS) open a door towards more dynamic and demand-responsive urban transportation. SAMS are expected to enhance efficiency by leveraging autonomous vehicle (AV) technology to optimize fleet operations and maximize vehicle utilization by repositioning idle vehicles, thereby reducing wait times and increasing the overall throughput of urban transport systems. In addition to efficiency, the equitable distribution of transportation services is emerging as a key concern for transportation policy and mobility planning. This study contributes to equity-focused operational strategies for SAMS by developing and evaluating a Reinforcement Learning (RL)-based AV repositioning strategy, with implications for mobility planning and policy design. The proposed model (AVR-GC) uses a reward function that incorporates the Gini coefficient of passenger waiting times to emphasize balancing service quality across zones. Our study analyzes the impact of this model through agent-based simulation experiments on the distribution of wait times and operational efficiency, comparing it against the AV Repositioning Baseline Approach (AVR-BA), which does not specifically target equity in its repositioning strategy. The results show that the AVR-GC model produces more uniform wait times across the service region during weekdays, indicating improved service equity. However, it also leads to longer average empty relocation distances, highlighting a trade-off between fairness and operational efficiency. The study also examines the robustness of these models under varying cancellation thresholds, defined as the maximum waiting time a passenger is willing to tolerate before canceling a request. Challenges in managing service variability persist, particularly under less restrictive conditions such as high cancellation thresholds. Discussing these aspects, this study aims to provide valuable insight to achieve an equitable operational framework for flexible mobility systems through fleet repositioning.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Towards equitable operation of shared-use autonomous mobility services via reinforcement learning for fleet repositioning

  • 期刊论文

  • 发表日期：

    2026年5月25日

查看全文
  • Haimanti Bala,
  • Monika Filipovska,
  • Archie J. Huang

摘要

Shared Autonomous Mobility Services (SAMS) open a door towards more dynamic and demand-responsive urban transportation. SAMS are expected to enhance efficiency by leveraging autonomous vehicle (AV) technology to optimize fleet operations and maximize vehicle utilization by repositioning idle vehicles, thereby reducing wait times and increasing the overall throughput of urban transport systems. In addition to efficiency, the equitable distribution of transportation services is emerging as a key concern for transportation policy and mobility planning. This study contributes to equity-focused operational strategies for SAMS by developing and evaluating a Reinforcement Learning (RL)-based AV repositioning strategy, with implications for mobility planning and policy design. The proposed model (AVR-GC) uses a reward function that incorporates the Gini coefficient of passenger waiting times to emphasize balancing service quality across zones. Our study analyzes the impact of this model through agent-based simulation experiments on the distribution of wait times and operational efficiency, comparing it against the AV Repositioning Baseline Approach (AVR-BA), which does not specifically target equity in its repositioning strategy. The results show that the AVR-GC model produces more uniform wait times across the service region during weekdays, indicating improved service equity. However, it also leads to longer average empty relocation distances, highlighting a trade-off between fairness and operational efficiency. The study also examines the robustness of these models under varying cancellation thresholds, defined as the maximum waiting time a passenger is willing to tolerate before canceling a request. Challenges in managing service variability persist, particularly under less restrictive conditions such as high cancellation thresholds. Discussing these aspects, this study aims to provide valuable insight to achieve an equitable operational framework for flexible mobility systems through fleet repositioning.