Preliminary Results from the Comparison of TFM Optimization Technique Outputs with Traffic Management Initiatives and Sensitivity Analyses

Abstract

Traffic Flow Management (TFM) manages demand at airport, airspace, and other National Airspace System (NAS) resources to maintain an efficient NAS-wide traffic flow, consistent with safety requirements. In the current air traffic management (ATM) system, the Air Traffic Control System Command Center (ATCSCC) traffic flow managers use decision support tools to identify areas of likely future demand-capacity imbalance and then use their operational experience to develop TFM initiatives such as Ground Delay Programs (GDPs) and Airspace Flow Programs (AFPs). There is no formal optimization framework in place. Moreover, decisions have to be made based on uncertain demand and capacity forecast information. As a result, most TFM delay decisions may be conservative. One promising TFM optimization algorithm was proposed by Bertsimas and Stock-Patterson in their 1990 seminal paper. This was an integer programming based optimization solution for computing optimum ground and airborne delays, while avoiding sector and airport capacity violations. Recently, NASA has developed methods to solve the integer programming problem in short times. Here we focus on the validation of this Bertsimas Stock-Patterson (BSP) solver by comparing its TFM controls against real-world Traffic-flow Management Initiatives (TMI). We also present preliminary results of the sensitivity of the BSP solver to the rolling horizon parameters (i.e., how often the optimizer is run, how far in the future it predicts flight trajectories to assess demand-capacity imbalances, and how far in the future it exercises control over departure and airborne delays). Moreover, as a part of this validation work we discovered a number of challenges related to the real-world implementation of the BSP TFM controls. We discuss these challenges in detail in this paper. I. Introduction and Motivation raffic Flow Management (TFM) regulates demand for the National Airspace System (NAS) resources using mechanisms such as gate delays, en route delays, and flight plan reroutes. This allocation process may use deterministic or probabilistic data and reasoning. The scope of the solution may span the entire NAS or it may be regional. The solution may be over shorter (tactical) time horizons or longer (strategic) time horizons, and user preferences may be incorporated. TFM controls used in the current-day air traffic system fall into two major categories. First, there are controls that are applied in the pre-flight phase, such as departure gate delays and flightplan reroutes. Second, there are controls that are applied in the en route phase, such as altitude capping, horizontal path-stretches and miles-in-trail restrictions. While current day TFM controls are effective, new algorithms for optimizing TFM controls are needed to keep delays at a manageable level in the future. Multiple research efforts have developed TFM algorithms to fulfill this need. Some have applied Integer