Rithwik Neelakantan, Digantara Research and Technologies Private Limited, India; Sivalinga Raja Shanmugam, Digantara Research and Technologies Private Limited, India; Ankit Agrawal, Digantara Research and Technologies Private Limited, India; Latha S, Digantara Research and Technologies Private Limited, India; Nistala Venkat Viashnavi Lakshimi, Digantara Research and Technologies Private Limited, India; Tanveer Ahmed, Digantara Research and Technologies Private Limited, India
Keywords: Pattern of life characterization, maneuver detection, Low Earth Orbit
Abstract:
Humanity’s utilization of space in the last century has been marked by significant advancements, from satellite launches to crewed missions to the Moon. The proliferation of satellites for communication, navigation, and scientific research has resulted in exponential increase in global accessibility, technological advancement and convenience. However, the increase in satellite deployment in recent times has highlighted the urgent need for a robust space operations infrastructure. Developing a comprehensive space infrastructure is essential to ensure the safety and sustainability of future space operations. A critical requirement for this development is the ability to characterize, synthesize and evaluate the pattern of life (PoL) of all resident space objects (RSO). A good starting point for this meticulous task would be the characterization of all active assets in the Low Earth Orbit (LEO).
The PoL characterization of a satellite in LEO is an intricate process demanding high fidelity data and robust data processing techniques. The inferences drawn from such an analysis can be subjective, depending on the requirements of the end user. The lack of ground truth data to test the reliability of the algorithms and the lack of commonality among different perspectives, requirements and definitions of PoL characterization demand a solution based on multiple perspectives and the capability to ingest and fuse data from multiple sources of varying fidelity. In this context, the current research proposes a dynamic and robust methodology to precisely characterize and synthesize PoL of satellites based on a multi-perspective multi-modal analysis, involving all possible aspects of an SDA technological chain. This involves processing, interpreting and correlating the following aspects: a) knowledge of the sensor and the output data type b) state information (position, velocity, attitude), its uncertainty and latency c) understanding of the limitations of the data processing technique from observations to states d) physical characteristics, propulsion system and payload characteristics of the object (if known) e) truth orbit data (if available) f) generating the maneuver history of the object, derive patterns & characterize anomalous behavior g) post processing to interpret/infer the purpose of the maneuvers etc. Because of the underlying modular approach, the fidelity required for the different aspects of the processing chain can be granular and coordinated. For example, if the confidence or latency in sensor observations are not adequate, the sensors can be tasked more enabling a system level improvement.
Towards building such an informed and data driven analysis module, this research employs a robust in-house multi-modal catalog of space objects and a high fidelity maneuver detection tool. The heuristic maneuver detection tool can ingest multiple data types, in multiple formats and performs a robust time series statistical analysis to characterize maneuvers. The fundamental concept employed is a sliding window analysis which iteratively computes the standard deviation of the specified parameter (such as semi-major axis, inclination etc.) within overlapping windows of specified length. Within each window, the process is iterative and it entails extracting the maximum value from the initial window until the size diminishes to a predefined threshold. At each step standard deviation is computed and compared with the modified window to evaluate the influence of the extracted value through ratio. Further the logarithm of ratio is computed to provide confidence level insights into relative changes within the window.
The above mentioned methodology can be extended for real-time detection of maneuvers. This process involves predicting the observations of the satellite from a known state and estimating the deviations of the actual sensor observations from the predicted observations. In the simpler case of a known object (assuming that the correlation between tracks is successful), the maneuver parameters (start, end times, magnitude, orbital change induced and deltaV spent) are deduced instantly when the maneuver is completed. In the case where the tracklet association is yet to be performed, a Multiple Hypothesis Testing algorithm is implemented and subsequently, the maneuver parameters are computed. The results of the real time maneuver detection can be made available to the stakeholders with acceptable latency limits, laying the foundation for informed decision making.
The PoL characterization of the satellite 40XXX in the duration from 2018 to 2020 is presented as a sample result. This satellite conducted five along-track maneuvers in the analyzed time frame. The temporal distribution of these maneuvers are random in nature, the least duration between maneuvers being 50.32 days and the most duration was 165.71 days. One of these maneuvers resulted in a change in semi-major axis behavioral trend, where a large change of about 7.741 km was imparted. From the maneuver characteristics, the satellite is deduced to have a chemical propulsion system. Further results will be included in the manuscript of the full paper.
The presented results in this abstract will be extended to the analysis of more satellites. The planned work also involves deriving the mission related parameters such as mass of fuel left, estimated mission duration, analysis of delta V budget for de-orbiting etc. The precise identification of the purpose of maneuver hence anomaly detection are also in the scope of the work.
The applications of the proposed technology development extend from the precise characterization of the behavior of adversary RSOs to the prediction of their future behavior. Through the multi-perspective approach, comprehensive evaluation of the SDA system level capabilities can be conducted and the system can be curated to generate more data or tune the analytics.
Date of Conference: September 17-20, 2024
Track: Satellite Characterization