The Satellite Pattern-of-Life Identification Dataset (SPLID) consists of a public challenge dataset and a private evaluation dataset. The public challenge dataset contains 500 simulated satellite trajectories, each operating under different mission objectives and equipped with various propulsion capabilities. These trajectories are generated using a high-fidelity astrodynamics simulator and aim to represent a diverse range of satellite behaviors. The private evaluation dataset includes additional data used to evaluate the performance of participants' models. This private evaluation dataset comprises both simulated satellite trajectories and historical satellite trajectories generated from VCM data and high-accuracy ephemerides provided by satellite operators.
The dataset encompasses astrometric data over six months at a two-hour temporal resolution, consisting of osculating orbital elements, geographic positions, and Cartesian states of the satellite in the J2000 inertial reference frame.
Each data point is accompanied by a list of expert-annotated time-stamped pattern-of-life (PoL) nodes. Each entry in the satellite PoL node list includes a TimeIndex, direction ("EW" for east-west and "NS" for north-south), node type, and propulsion type used while in this behavioral mode. The TimeIndex corresponds to the row index of the astrometric data for that specific node.