Conformance Monitoring in Air Traffic Control

Conformance monitoring in Air Traffic control is an approach that is used to monitor and detect any deviations of an aircraft's flight path from the assigned flight plan, that might compromise safety or efficiency. Currently, the task of conformance monitoring is performed by air traffic controllers based on comparisons between observed aircraft positions from radar surveillance and the expected aircraft positions based on clearances or flight plans. If the differences between the observed and expected positions exceeds some allowable deviations, non-conformance can be determined. Due to limited resources available to the air traffic controllers, this method does not give good performance. The conformance monitoring task would be even more difficult in future ATC operations. Under the NextGen, a larger number of aircraft are expected to operate with reduced separation thresholds between aircraft within a given airspace. The new concept of operation also allows aircraft the flexibility of changing flight routes (or flight plans) in response to changing conditions. Furthermore, different aircraft would have very different navigation capabilities due to different levels of equipping. Under such complex scenarios in future ATC operations, it would be important to have a conformance monitoring tool for monitoring the aircraft movements.

We have proposed a conformance monitoring algorithm, based on a fault detection concept, with the objective of addressing some of the challenges that could be encountered in current and future ATC operations. Our work is motivated by the work of Reynolds and Hansman [Tom G. Reynolds and R. John Hansman, Investigating conformance monitoring issues in air traffic control using fault detection techniqes, Journal of Aircraft, 44(2):1307-1317, 2005.] which discusses the challenges of conformance monitoring in various different ATC operational scenarios. The general idea of the proposed conformance monitoring scheme is illustrated in Figure 4. Here, the actual system represents the dynamics of an aircraft and the measurements are its observed trajectory. The model of the system describes the expected behavior of a conforming aircraft. In practice, for ATC applications, the command input is unknown. However, flight plan and clearance information can be used as input to the model. The conformance monitoring algorithm is implemented as a fault detection function which consists of a residual generation scheme and a decision making scheme. The residual generation scheme generates a vector known as the residual which is used to test for conformance or non-conformance by the decision making scheme. The contributions of our work are (1) the proposal of an effective residual generation scheme that generates a residual with zero mean when the observed aircraft trajectory conforms to our aircraft dynamics model, and (2) the derivation of an approximate statistical characteristics of the residual which facilitates the implementation of the statistical decision making algorithm. Many conformance monitoring algorithms models that the aircraft follows the planned flight path without any navigation errors. This is not true in practice and it results in higher detection delays in the regions where there are high navigation uncertainties, such as during a turn in the vicinity of a waypoint. Our aircraft dynamics model, described using the SLHS, is able to describe the aircraft trajectory deviations due to navigation uncertainties. As a result, our conformance monitoring algorithm yields smaller detection delays compared with other conformance monitoring algorithms (see Figure 6).

Figure 5: Conformance monitoring based on a model-based fault detection concept.

Figure 6: Comparison of performance of proposed conformance monitoring algorithm with that of the CR algorithm proposed by Reynolds and Hansman. The ADS-B data comes from the Automatic Dependent Surveillance-Broadcast (ADS-B) system which has a sample time of 1 sec. The radar data comes from an ATC surveillance radar with a sample time of 6 sec. The CR algorithm only has result based on ADS-B data.