Authors: Monika Filipovska, Michael Hyland, Haimanti Bala

Published on: October 16, 2022

Impact Score: 8.07

Arxiv code: Arxiv:2210.08659

Summary

• What is new: Introduces an integrated system-agent repositioning (ISR) approach for shared-use autonomous mobility services (SAMS) fleet repositioning, utilizing reinforcement learning without the need for explicit demand forecasting.
• Why this is important: The challenge of optimally positioning idle fleet vehicles in SAMS to meet future demand, impacting service quality and efficiency.
• What the research proposes: The ISR approach formulates fleet repositioning as a Markov Decision Process and uses reinforcement learning for scalable strategy development in response to evolving demand patterns and optimization-based passenger-to-vehicle assignment.
• Results: Substantial reduction in passenger wait times by over 50% compared to the joint optimization approach, demonstrating ISR’s effective response to evolving demand patterns and its transferability under various conditions.

Technical Details

Technological frameworks used: Markov Decision Process, Reinforcement Learning

Models used: Integrated System-Agent Repositioning (ISR), Externally Guided Repositioning (EGR), Joint Optimization (JO)

Data used: New York City taxi data, agent-based simulation tool

Potential Impact

Urban mobility services, autonomous vehicle companies, and public transportation agencies could be significantly impacted. Positive disruption in terms of operational efficiency and service quality improvement is expected.

Want to implement this idea in a business?

We have generated a startup concept here: FleetFlow.