Research Analyzer
← Back ICRA 2026

Large-Scale Autonomous Vehicle Fleet Management

Timothy Mulumba

PDF

AI summary

Key figure (auto-extracted from paper)
A hierarchical optimization framework jointly coordinates dispatch, charging, and facility ingress for city-scale autonomous fleets, significantly improving service coverage and wait times while preventing charger overuse.
autonomous mobility-on-demand fleet management hierarchical optimization mixed-integer programming charging infrastructure constraint-aware control

Problem

Prior autonomous mobility-on-demand formulations typically handle trip assignment or long-horizon charging in isolation, failing to jointly enforce hard physical constraints like limited charger capacity and facility ingress limits across multiple operational time scales.

Approach

The authors propose a two-tier system where a coarse-grained Deployment model plans day-long vehicle flows between the city and facilities, which then guides a fine-grained Summoning mixed-integer program that assigns specific vehicles while respecting real-time capacity and travel constraints.

Key results

  • Unified hierarchical framework enforcing hard charging and facility ingress constraints
  • Interoperable coarse-grained Deployment and fine-grained Summoning optimization workflows
  • City-scale evaluation showing higher demand coverage and shorter wait times than greedy baselines
  • Near-zero charger overuse (≈1–2%) and a concrete RL training protocol with MIP safety shielding

Why it matters

Enables safe, scalable, and grid-friendly deployment of large autonomous ride-hailing fleets by bridging the gap between strategic resource planning and tactical dispatch under real-world physical constraints.

Abstract

We present a hierarchical framework for city-scale autonomous ride-hailing that integrates vehicle prepositioning, request matching, charging, and facility ingress. A fine-grained mixed-integer program (MIP) coordinates prepositioning and matching on short horizons, while a coarse-grained Deploy- ment+Summoning decomposition enforces charger/parking ca- pacities at scale. On ride-hail traces, the method increases coverage and reduces wait relative to greedy and decoupled baselines, while keeping charger overuse near zero (≈1–2%) under rolling-horizon execution. We detail boundary-condition handling for 24/7 operations and specify a concrete RL train- ing/validation protocol for a constraint-aware hybrid in which learned policies act tactically under a MIP-based safety shield.

Index terms

Autonomous Vehicle Navigation Multi-Robot Systems Optimization and Optimal Control

Related papers