Detection of Satellite Manoeuvres Using Non-Linear Kalman Filters on Passive-Optical Measurements

Abstract

As part of an ongoing effort to build and maintain a data base for Space Situational Awareness, we have been developing an algorithm that employs a non-linear Kalman Filter to detect satellite manoeuvres. This methodology works directly on the astrometric angle measurements derived from passive-optical telescope observations without the need to run an orbit determination step first. In this study, we analyze the performance of this algorithm using large data sets of synthetic observations, and use it to detect and characterize several manoeuvres performed by geostationary satellites. In order to assess the capabilities and limitations of this method, we first created a large set of synthetic observations based on precisely known orbits of Galileo satellites. With that in hand, we study the effects of varying different properties of the data, such as the noise level, manoeuvre magnitude, and manoeuvre direction. As a second step, we apply the manoeuvre detection algorithm to real astrometric observations of two geostationary satellites obtained with optical telescopes of the SMARTnet sensor network, including observations during the launch and early orbit phase. Like our simulated observations, our real-world data set comprises manoeuvres of very different magnitudes and directions. In some cases, the true manoeuvre details were known to us, so we could verify our findings. We detect a number of manoeuvres in our observations, and in one case, we also accurately determine the manoeuvre epochs and ∆v-components. This is done by means of a conjunction analysis, during which we calculate the collision probability between two tracks propagated forward and backward, respectively. We then determine the manoeuvre epoch and ∆v-components at the time of maximum collision probability. We show that with this method we can determine the manoeuvre epoch to within a few seconds, and the ∆v-components to an accuracy at the cm/s-level.