Optimization-Based Simulation of Air Traffic Flow Management Within the Decentralized ASEAN Region

Air Traffic Flow Management (ATFM) tactical measures, such as imposing acceptance rates and calculated take-off times, rely on information sharing and collaboration across Air Traffic Control sectors. In decentralized systems, such as those in Southeast Asia, this collaboration poses significant challenges, often limiting tactical measures to implementation within individual Flight Information Regions (FIRs). This paper presents the development of simulation capabilities to analyze such decentralized ATFM systems, where FIRs operate independently. We propose a comprehensive suite of mathematical models and algorithmic frameworks specifically designed for decentralized ATFM simulations. These models incorporate essential airspace constraints, including airport and airspace capacity limitations, as well as aircraft separation requirements. The proposed algorithms operate within each independent FIR, leveraging local information at each time step and applying a rolling-window scheme to optimize ATFM decisions, such as speed adjustments and airborne holding. The objective is to optimize the Target Time Over (TTO) for flights at specific waypoints within each FIR while minimizing deviations from planned flight times. We evaluate the proposed algorithms under two different airspace capacity scenarios within the Southeast Asian region, simulating a three-hour period with a two-hour look-ahead horizon advanced in five-minute rolling steps. Our results demonstrate that the algorithms generate conflict-free solutions (i.e., solutions satisfying all node-level flow-separation constraints) for a network-wide decision window, the fastest within approximately 10 s, so that a three-hour simulation with five-minute steps can be completed in about five minutes. Furthermore, we illustrate the impact of reduced capacity periods, highlighting how our models facilitate insightful analysis of delay growth due to capacity restrictions. Finally, we compare the computational trade-offs and effectiveness in minimizing airborne delays among Exact Methods (EMs), Gradient Descent Ascent (GDA), and Simulated Annealing (SA).