Home > News & Events > MITRE Publications > The Edge >

To the untrained eye, a graphical display of air traffic trails looks like a tangled, confusing web. In reality, airspace usage is orderly, managed in a systematic way by the air traffic management services of the Federal Aviation Administration (FAA). Air traffic, especially in complex, congested airspace, is structured along predefined routes. Sometimes, however, air traffic controllers allow certain flights to fly off of these predefined routes in order to expedite traffic or to satisfy pilots’ requests. The push to provide airspace users greater flexibility, plus the dramatic increase in air traffic volume, are making it more challenging for planners and designers to determine airspace usage patterns.

Airspace planners and designers need to understand airspace usage patterns in order to define boundaries of sectors (the airspace units within which controllers work) that facilitate air traffic movement yet do not overwork air traffic controllers. To assist them, MITRE’s Center for Advanced Aviation System Development (CAASD) has developed a new approach to determining airspace usage patterns, a data mining system called the Dominant Flow Detector (DFD). The system is an outgrowth of MITRE-Sponsored Research conducted last fiscal year that focused on understanding interrelationships among air traffic flows in the congested Northeast United States.

Unstructured Air Traffic Poses Challenges for Planners
Structured air traffic facilitates expeditious movement in high-density areas. Controllers can handle more operations when they know sector entry and exit points together with flow merge and altitude transition points.

Unstructured traffic, by contrast, does not follow established patterns through airspace. It affords pilots flexibility to pursue their own preferences for route, altitude, and speed. Increased flexibility is a feature of Free Flight, an industry/government initiative promoting greater freedom and autonomy for the user as well as greater collaboration between airspace users and managers.

The desire for increased flexibility on the part of the pilot makes it imperative that sectors are structured to accommodate the flow and patterns of traffic efficiently. But before planners and designers can implement changes in the airspace boundaries that define sectors, they need to understand dominant flows, and be able to discriminate between largely structured and largely unstructured airspace volumes.

The DFD Eliminates Manual Calculations
That’s where the DFD comes in. By employing a statistical technique called cluster analysis, plus data manipulation logic to refine the cluster analysis output, the DFD system eliminates the need to manipulate airway charts and voluminous operational data manually.

Here’s how the DFD operates. For a specified volume of airspace, a sample of operational data is examined. The entry and exit points of flights that penetrate the airspace are recorded. The set of all such flights is considered the data set to be analyzed. The cluster analysis algorithm determines groupings within the data set according to the similarity of the entry and exit points for members of the set.

The challenge for the DFD is that airspace sector sizes and air traffic patterns and densities vary widely throughout the National Airspace System. The DFD needs to provide good results in spite of these varying conditions. But the clustering algorithm “knows” nothing about airspace, flights, or even horizontal and vertical coordinate systems. It can only do two simple things: measure the similarity of two observations and produce groupings of those that are sufficiently alike. It is therefore necessary for the DFD to examine the output of the cluster analysis and make sense of it in the context of air traffic flows.

Graphical depictions in the figures illustrate some of the results achieved by the DFD. Working in two dimensions, the DFD will consider Figure 1a air traffic flight trails in a sector and produce Figure 1b flows. Figure 1a shows all the tracks, with dominant flows obscured by unstructured traffic. By eliminating approximately 40 percent of tracks that do not fall into an entry/exit point cluster, we can easily see the dominant traffic flows in Figure 1b. (Figures 1a and 1b are taken from experiments associated with developing the DFD. The sector illustrated is in the Washington Air Route Traffic Control Center (ARTCC) airspace; the data are from 10/23/98.)

The X/Y scaling shows relative distances in nautical miles. In Figure 1b, we see that the DFD has filtered out flights that are “singletons” or one-off actions by air traffic controllers. Also filtered away are anomalous observations, since a few trails can be seen beyond the sector boundary in Figure 1a. What remains are coherent and easily discernable air traffic flows through the sector.

3-D View Portrays Altitude Transitions
Because altitude transitions in an airspace volume are important information for planners, the DFD has been developed to operate in three dimensions as well as two. Figure 2a shows a volume of airspace with air traffic flight trails through it. (The data were taken in support of sponsored research from the Atlanta ARTCC airspace, also on 10/23/98.) One might guess that there are several dominant flows through the airspace. But that is difficult to discern, since the figure is busy with “noise,” i.e., air traffic not part of a dominant flow. The DFD yields the results shown in Figure 2b, in which dominant flows are color-coded. Five flows are displayed, some with altitude transitions, others in level flight.

Big-Picture Applications
It is expected that the DFD will have continuing utility in airspace analysis work for the FAA. For example, the goal of one current FAA research project is to understand the airspace usage and dominant air traffic flows in a 10-state region in the Southeast United States. This “big-picture” examination, an initial application of the DFD, will probably result in recommendations on airspace boundaries that are more consistent with dominant flows. Such boundaries would facilitate existing flow patterns while keeping the workload of individual air traffic controllers manageable.

---

The contents of this material reflect the views of the author and/or the Director of the Center for Advanced Aviation System Development. Neither the Federal Aviation Administration nor the Department of Transportation makes any warranty or guarantee, or promise, expressed or implied, concerning the content or accuracy of the views expressed herein.

---

For more information, please contact Jim DeArmon using the employee directory.

---