Simulation Framework for Rebalancing of Autonomous Mobility on Demand Systems

Marczuk Katarzyna A.; Soh Harold S.H.; Azevedo Carlos M.L.; Der-Horng, Lee; Emilio, Frazzoli · 2016 · DOAJ

DOI: 10.1051/matecconf/20168101005

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the operational challenges of Autonomous Mobility on Demand (AMOD) systems, specifically focusing on vehicle rebalancing to ensure service availability and efficiency. While shared-use mobility offers sustainable alternatives to private car ownership, unbalanced fleets often lead to service unavailability during peak demand. The authors propose that autonomous vehicles can mitigate these issues by reducing operating costs and enabling automated redistribution. The study aims to evaluate the impact of different rebalancing policies on AMOD system performance, measured by fleet size requirements and customer waiting times. The methodology employs a simulation-based approach using SimMobility, an agent-based microscopic simulation platform. The authors developed an AMOD Controller to manage fleet operations, including facility location, passenger assignment, routing, and rebalancing. The study compares three rebalancing strategies: (i) no rebalancing (baseline), where vehicles move only when assigned to customers; (ii) offline rebalancing, which uses historical data and linear programming to determine optimal initial vehicle distribution and rebalancing counts before simulation; and (iii) online rebalancing, which dynamically optimizes vehicle movements during the simulation based on real-time network conditions and predicted demand. The simulation was conducted on a 56 km² network in Singapore’s Central Business District, utilizing demand data derived from the 2012 Household Interview Travel Survey. The offline model used a 15-minute rebalancing interval, while the online model used a 1-hour interval. Results indicate that rebalancing significantly improves system efficiency. To achieve an average customer waiting time below 10 minutes, the no-rebalancing policy required a fleet of 35,000 vehicles. In contrast, the online rebalancing policy required only 25,000 vehicles, representing a 28% reduction in fleet size. The offline model estimated a minimum fleet size of 24,216 vehicles, though it resulted in slightly longer waiting times (11.62 minutes) compared to the online model for similar fleet sizes. The authors attribute discrepancies between offline and online results to the offline model’s reliance on time-invariant travel times and immediate service assumptions, whereas the online model accounts for dynamic congestion and customer queues. The significance of this work lies in demonstrating that automated rebalancing is critical for the viability of AMOD systems, allowing for substantial reductions in required fleet sizes and associated infrastructure costs like parking. The findings suggest that online rebalancing offers a more realistic and efficient approach for real-time operations compared to static offline planning. The authors conclude that future work should focus on hybrid models combining the strategic benefits of offline planning with the dynamic responsiveness of online control, as well as incorporating demand management strategies such as dynamic pricing.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• driverless ads