Cameron Webb, Defence Science and Technology Group; Mark Rutten, InTrack Solutions; Zoe Wolff, Defence Science and Technology Group
https://doi.org/10.64861/BLXW1788
Keywords: GEO, Satellite Manoeuvre, Satellite Maneuver, Pattern of Life, TLE, Anomaly Detection
Abstract:
Space Domain Awareness (SDA) aims to achieve a comprehensive understanding of activities occurring on orbit. Satellites with highly regular station-keeping manoeuvres, such as Geostationary Earth Orbit (GEO) satellites, establish patterns of life from which deviations may indicate operational changes, mission events, relocation campaigns or other behaviours that may be of interest to SDA analysts. Recent studies have explored a range of factors for establishing satellite patterns of life, including orbital location data, light curves, manoeuvre frequency, and other behavioural indicators. This paper presents a proof of concept methodology that employs a manoeuvre-based pattern of life algorithm to predict GEO satellite station-keeping activities and flag candidate anomalous manoeuvres, supporting first-pass alerting of SDA analysts to events of interest.
The methodology uses publicly available Space-Track Two-Line Element (TLE) data to infer station-keeping activity and construct manoeuvre-based patterns of life for individual GEO satellites. TLE derived orbital parameters are processed with a Kalman filter-based change-detection algorithm to extract candidate East-West and North-South manoeuvres. These candidate times are then used to estimate satellite-specific longitude-control behaviour, inclination-control behaviour and manoeuvre intervals, capturing uncertainty in each parameter. A recursive algorithm has been developed to characterise routine manoeuvre timing, which allows the behaviour pattern to be updated as each elset is received. These models of behaviour can be used to predict future manoeuvre windows and expected operating bounds, allowing newly detected manoeuvres to be assessed against the learned pattern of life. Manoeuvres occurring outside the predicted temporal or spatial confidence bounds are flagged as candidate anomalies.
The paper demonstrates this framework using 30 operational GEO satellites, employing chemical propulsion systems and 1.5 years of historical orbital data for testing. The evaluation also includes additional deliberately challenging cases, such as satellites with irregular station-keeping behaviour, noisy or irregular TLE histories, and low-thrust electric propulsion, to assess the limits of the proposed approach. The approach is best suited, within this study, to GEO satellites using chemical propulsion for both East-West and North-South station-keeping, where manoeuvres are more readily observable in TLE-derived orbital behaviour. The performance of the manoeuvre detection and pattern of life generation algorithms are evaluated through comparison with owner/operator manoeuvre data and expert analysis. Manoeuvre prediction and anomaly detection are further investigated and benchmarked through selected test cases to emphasise specific satellites and time periods.
Overall, the results demonstrate the feasibility of using a manoeuvre-based pattern of life modelling method to support first-pass candidate anomaly screening for GEO satellites. Routine station-keeping behaviour is shown to be sufficiently repeatable, to enable probabilistic prediction of future manoeuvre windows and automated flagging of candidate anomalous manoeuvres. However, performance remains dependent on TLE quality, station-keeping regularity, propulsion type, and parameter selection. These limitations motivate future work using higher fidelity tracking data, broader validation datasets, and further refinement of the manoeuvre detection and prediction algorithm.
Date of Conference: September 15-18, 2026
Track: Space Domain Awareness