GEO spacecraft maneuver detection based on causal inference

Abstract

Geosynchronous (GEO) spacecraft maneuver detection is deemed crucial for space domain awareness (SDA) as it helps to maintain catalogs of high value resident space objects (RSOs). However, several difficulties in resolving maneuver time and requirements for historical data quality still exist for the current research. This paper proposes a method based on causal inference for accurate and timely detection of the maneuver using real-time observational data from optical sensors. Firstly, the residual information of time series observations is obtained through counterfactual reasoning. Secondly, according to the Gaussianity of the optical sensor observation noise, a structural causal equation (SCE) of the residuals of time series observations is derived. Lastly, the maneuver is abstracted as a soft intervention, and the minimum average treatment effect (MATE) is proposed to estimate the effect of the maneuver and used as a feature parameter to identify the GEO spacecraft maneuver. After experimental analysis, the detection rate and detection timeliness of the proposed method reach more than 90%, a significant improvement compared with the conventional χ2, skewness, and kurtosis test methods.