Anomaly Detection for Spacecraft by Estimating Parameters with Particle Filter

This paper proposes a method of early spacecraft anomaly detection by simultaneously estimating its states and parameters. We applied an extended particle filter algorithm in order to estimate not only states but also parameters. In this method, we incorporated artificial evolution of parameters and kernel smoothing of parameters into the ordinary particle filter algorithm. Each parameter is related to each state of the spacecraft components, so we can understand what is happening in the spacecraft by finding out parameters’ changing signs. We tested the algorithm on a simulation of spacecraft attitude motion.