Planning UAM network under uncertain travelers’ preferences: A sequential two-layer stochastic optimization approach

Urban Air Mobility (UAM) holds significant promise for enhancing travel efficiency and improving regional accessibility. However, policymakers face a fundamental challenge: infrastructure planning decisions must often be made before demand is known. This study develops a single-stage stochastic optimization framework with sequential decision layers that mirrors real-world planning constraints. It allows agencies to determine vertiport locations and trip allocations before individual mode choices are realized, incorporating behavioral uncertainty via discrete choice modeling and Monte Carlo simulation. To ensure computational tractability at realistic scales, an improved greedy algorithm (GRD-U) is introduced and benchmarked against established heuristics. Experiments on synthetic instances show that cost-saving potential is greatest in larger regions with low road connectivity, as well as unicentric or dispersed demand patterns. A real-world case study in the Munich Metropolitan Area confirms the framework’s applicability, demonstrating notable improvements in generalized travel cost savings, demand coverage, and accessibility compared to existing siting strategies. A sensitivity analysis highlights how UAM performance responds to changes in operational parameters, such as cruise speed, pricing strategies, and vertiport quantity. The framework offers a transparent and behaviorally grounded tool for early-stage UAM planning. It enables public agencies to anticipate demand patterns under uncertainty, weigh trade-offs between investment scale and system performance, and align infrastructure planning with equity and efficiency goals. These contributions provide practical decision support for cities navigating the complexities of UAM deployment.