Fleet Size and Spill for UAM Operation under Uncertain Demand

TL;DR

This paper addresses fleet sizing for Urban Air Mobility under uncertain demand by coupling a stochastic demand generator, built from real-world data, with two integer-programming models that yield (i) a zero-spill fleet size and (ii) spill-minimizing flight schedules and charging policies. The methodology captures both day-to-day and within-the-day demand variation and models non-linear charging times across discrete state-of-charge levels, enabling realistic routing and energy management over a two-vertiport network. Key findings show that spill is relatively inelastic to fleet size, with the driving factor being imbalance in directional demand; larger fleets reduce spill but the benefit diminishes, particularly under strong demand imbalance. The work offers practical insights for UAM operators on how to size fleets and design charging and scheduling policies under demand variability, with implications for operability, capital planning, and resilience in urban air networks.

Abstract

Variation and imbalance in demand poses significant challenges to Urban Air Mobility (UAM) operations, affecting strategic decisions such as fleet sizing. To study the implications of demand variation on UAM fleet operations, we propose a stochastic passenger arrival time generation model that uses real-world data to infer demand distributions, and two integer programs that compute the zero-spill fleet size and the spill-minimizing flight schedules and charging policies, respectively. Our numerical experiment on a two-vertiport network shows that spill in relatively inelastic to fleet size and that the driving factor behind spill is the imbalance in demand.

Figures (8)

• Figure 1: Analytical Framework
• Figure 2: An example of 100 realizations of autoregressive passenger arrival time generation using the flight schedule of a major U.S. airport on 1/1/2019. The autoregressive coefficient is set to $\alpha=0.7$. The realized passenger arrival process vary around the expected hourly rate $\Lambda^{0}_{d,h,v}$ and the gap between the realized trajectories and the expected rate is due to the addition and subtraction of lead and lag time.
• Figure 3: Distribution of daily realized number of passengers across different demand levels ($ADD$)
• Figure 4: Aggregated hourly flight demand on one realization of passenger arrivals after the conversion from passenger arrival to flight demand using the 5-minute dispatch policy.
• Figure 5: Distribution of zero-spill fleet size computed by the fleet sizing optimization model for different sets of demand model parameters.