Ryo Kato, NEC Aerospace Systems, Ltd.; Takahito Sakaue, NEC Aerospace Systems,Ltd.; Daiki Mori, NEC Corporation; Makoto Tanaka, NEC Aerospace Systems, Ltd.; Taichiro Sano, NEC; Katsuaki Taya, NEC Corporation; Masatoshi Ebara, NEC Corporation; Ryosuke Nakazawa, NEC; Jun Yoshida, NEC; Ryosuke Togawa, NEC Corporation
Keywords: Machine Learning, Deep Learning, Metric Learning, Maneuver, SSA/SDA
Abstract:
Recently, there has been increased emphasis on satellites used for aspect of both our life and security. NEC operates remote sensing satellite ASNARO2, which was developed in-house. On the other hand, an increase in the number of space objects (e.g. satellite constellation or debris), anti-satellite weapon (ASAT), and such notable objects have led to the increasing risk of a conjunction when outer space is utilized. To utilize the outer space safely, it is necessary to manage the detailed orbits and detect truly critical conjunctions of space objects. There are a considerable number of several conjunctions in operational satellites; dozen conjunctions which we received from Combined Space Operations Center (CSpOC) as a Conjunction Data Messages (CDMs) have occurred in ASNARO2 over a year. Currently, NEC has managed orbit information of space objects using the SSA system called NEC ComSpOC, which is based on observation data of some optical telescopes and radars. In conducting the conjunction analysis, a SSA system must keep the determination accuracy of the space objects orbit, which is called Orbit Determination Confidence (OD confidence), high. In order to keep it high, it is necessary to capture maneuvers of space objects. NEC ComSpOC has leveraged the algorithm, however, NEC ComSpOC may overlook some maneuvers of space objects. In this case, an operator manually determine total delta V and when the maneuver was conducted. Due to the lack of human resources associated with increased space objects, it is difficult to manually determine maneuvers and keep OD confidence high. Therefore new tools and frameworks are needed to solve these operational problems to improve the capability of Space Situational Awareness (SSA) and Space Domain Awareness (SDA).
In this study we applied the model NEC developed for detecting maneuvers from two line elements (TLEs) of the notable objects of Geostationary Orbit (GEO) to NEC ComSpOC operations and evaluated the validity of the model.
The majority of previous studies on maneuver detection from TLEs are based on the stable satellite, which is assumptions with detailed maneuver history available and regular maneuvers. For this assessment, this study leverage the model NEC developed, which model obtains embedding of similarity between data points considering temporal information. This model is composed of the combination of time series and metric learning using Deep Neural Network (DNN). Metric Learning has been successfully used in various field, including object identification, natural language processing, speech recognition and medical diagnosis. Metric Learning learns data similarity on a distance basis and can perform similarity discriminations on unknown data. Furthermore, metric learning has achieved remarkable results in recent years when combined with DNN. This model learns mapping from TLEs to embedding in feature space, where similar TLEs are located nearby and dissimilar ones is far away, based on the unsupervised manner. To create the maneuver detection model in unsupervised manner, the model training is based on the labeled data using pseudo-labeling based on a distance.
In this study, our model trained with normal values and calculated the anomaly score in the TLEs of GEO objects which contains with high frequency maneuvers, especially those with significant differences from station keeping maneuver, whereas with scarcely maneuvers. As a results, it did partially substantiate the effectiveness of the model NEC developed for GEO objects and the feasibility of maneuver detection of notable objects based on TLEs. Furthermore, our model was partially successful in detecting the maneuver which NEC ComSpOC could not detect.
These findings could be used to help SSA/SDA operation. Further research could also be conducted to determine the effectiveness and feasibility of Low Earth Orbit (LEO) objects.
Date of Conference: September 19-22, 2023
Track: Machine Learning for SDA Applications