Deep RL-Based Optimization of AEV Fleets for Efficient Urban Mobility

• Grant number: 25/12288-6
• Support Opportunities: Scholarships abroad - Research Internship - Master's degree
• Start date: December 01, 2025
• End date: May 31, 2026
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Fabio Kon
• Grantee: Gustavo Henrique Santos Rodrigues
• Supervisor: Jakob Puchinger
• Host Institution: Instituto de Matemática e Estatística (IME). Universidade de São Paulo (USP). São Paulo , SP, Brazil
• Institution abroad: École Centralesupélec, France
• Associated to the scholarship: 24/18526-3 - Dynamic Management of Last-Mile On-Demand Transit: Designing Transferable Models for Deep Reinforcement Learning Algorithms, BP.MS

Abstract

The increasing demand for efficient and sustainable urban mobility solutions moves us towards new paradigms, such as using Autonomous Electric Vehicles (AEVs). Managing fleets of AEVs requires advanced optimization techniques, but these do not work well for large-scale deployments in dynamic environments. This research project evaluates the potential of Deep Reinforcement Learning (Deep RL) to address key operational challenges in AEV fleet management for On-Demand Transit (ODT) feeder services. We will use Deep RL-based strategies to generate optimal strategies for vehicle dispatching and battery recharging, where the actions will be to choose a dispatch direction, wait at a transit hub, or perform a partial or full battery recharge. The performance of implemented Deep RL algorithms will be benchmarked against traditional optimization techniques.