[Paper Review] Trajectory Clustering and an Application to Airspace Monitoring

Maxime Gariel, Ashok N. Srivastava+1 more|arXiv (Cornell University)|Jan 27, 2010

Air Traffic Management and Optimization33 references17 citations

TL;DR

This paper proposes a data-driven framework for real-time airspace monitoring using trajectory clustering to identify nominal flight patterns and detect deviations. It introduces a complexity metric based on Shannon entropy to quantify air traffic controller workload, enabling proactive flow management by detecting non-compliant aircraft in terminal airspace.

ABSTRACT

This paper presents a framework aimed at monitoring the behavior of aircraft in a given airspace. Nominal trajectories are determined and learned using data driven methods. Standard procedures are used by air traffic controllers (ATC) to guide aircraft, ensure the safety of the airspace, and to maximize the runway occupancy. Even though standard procedures are used by ATC, the control of the aircraft remains with the pilots, leading to a large variability in the flight patterns observed. Two methods to identify typical operations and their variability from recorded radar tracks are presented. This knowledge base is then used to monitor the conformance of current operations against operations previously identified as standard. A tool called AirTrajectoryMiner is presented, aiming at monitoring the instantaneous health of the airspace, in real time. The airspace is "healthy" when all aircraft are flying according to the nominal procedures. A measure of complexity is introduced, measuring the conformance of current flight to nominal flight patterns. When an aircraft does not conform, the complexity increases as more attention from ATC is required to ensure a safe separation between aircraft.

Motivation & Objective

• Address the challenge of increasing air traffic complexity in terminal airspace by automating the detection of non-standard flight operations.
• Develop unsupervised trajectory clustering methods to learn nominal flight procedures from radar data without relying on flight plans or ATC intent.
• Create a real-time monitoring tool, AirTrajectoryMiner, to assess conformance of current flights to learned nominal procedures.
• Introduce a novel complexity measure based on entropy to quantify controller workload and support flow management decisions.
• Enable application of the framework to diverse airspace configurations and other domains such as commercial aviation or ground fleet tracking.

Proposed method

• Use unsupervised clustering based on waypoint identification to detect common air routes from radar tracks in terminal areas.
• Apply principal component analysis (PCA) to resampled trajectories to extract dominant flight pattern shapes.
• Define nominal trajectories as clusters of similar flight paths derived from historical radar data, excluding visual flight rules (VFR) traffic.
• Compute a complexity metric using Shannon entropy: $ C = I_{SFO} + I_{ar{SFO}} $, where $ I_{SFO} $ and $ I_{ar{SFO}} $ represent entropy of aircraft conforming and deviating from nominal procedures at SFO.
• Estimate probabilities: $ p(OK|SFO) = \frac{n_{OK,SFO}}{n_{SFO}} $, and $ p(\bar{OK}_i|SFO) = \frac{1}{n_{SFO}} $ for each outlier, assuming independence.
• Implement AirTrajectoryMiner as a real-time tool that updates complexity every 15 seconds, using instantaneous aircraft counts and conformance status.

Experimental results

Research questions

• RQ1 How can nominal flight procedures be automatically learned from radar data without flight plan or intent information?
• RQ2 What trajectory clustering techniques are effective for identifying standard arrival patterns in terminal airspace with high variability?
• RQ3 How can a real-time measure of airspace complexity be derived from conformance to nominal procedures?
• RQ4 To what extent does the proposed complexity metric correlate with controller workload and traffic management needs?
• RQ5 Can the framework be generalized to other airspace configurations or transportation domains beyond aviation?

Key findings

• The trajectory clustering methods successfully identified standard arrival patterns at SFO using only radar data, without relying on flight plans or controller vectors.
• The PCA-based clustering method demonstrated robustness in capturing dominant flight path shapes despite variations in trajectory length and sampling rate.
• The proposed complexity metric $ C $ increases with the number of non-compliant aircraft and total traffic volume, reflecting higher controller workload.
• When no aircraft were detected as outliers, the complexity metric was zero, confirming its sensitivity to deviations from nominal operations.
• The tool's real-time update capability (every 15 seconds) enables timely detection of increasing complexity, supporting proactive traffic flow management.
• The framework is adaptable to other airports or transportation systems, such as GPS-equipped truck fleets, due to its unsupervised and data-driven nature.