Atmospheric drag uncertainty quantification for orbit determination and propagation via Stochastic Consider Parameters

• Uncertainty quantification method to improve covariance realism in Space Traffic Management and Space Situational Awareness services. • Stochastic model proposed to characterize the atmospheric drag aleatoric behaviour, based on linearization theory in batch least-squares estimation but including the impact of stochastic noise with correlation. • Validation and results for multiple resident space objects obtained with real measurements from the Spanish Space Surveillance Radar (S3TSR) to demonstrate operational applicability. • Covariance realism metrics used to assess the performance and realism improvement in a real space environment. The efficiency and sustainability of spacecraft operations is a growing challenge due to the accelerated increase of the space objects population. Thus, the quality of Space Traffic Management (STM) and Space Situational Awareness (SSA) services is essential to ensure the sustainability of the space environment. The quality of such services relies not only on an accurate knowledge of the Resident Space Object (RSO) state, but also on its associated uncertainty. However, there is a lack of accurate and cost-effective methodologies for Uncertainty Quantification (UQ) in the context of SSA, where a large number of objects are maintained in the catalogues. The atmospheric drag is one of the largest sources of uncertainty in Low Earth Orbits (LEO). Stochastic models have been widely proposed in the literature to represent its aleatoric nature. However, the introduction of stochastic dynamics increases the complexity of orbit propagation and determination. On the other side, classical implementations to characterize the uncertainty from dynamical models in batch least-squares orbit determination such as the consider parameters theory fail to represent the stochastic nature of the atmospheric drag uncertainty, despite maintaining a tractable level of complexity. This work presents the validation with real data of the Stochastic Consider Parameters (SCP) model, a methodology developed for uncertainty quantification via covariance estimation applied to batch least-squares orbit determination and propagation that allows considering the effect of stochastic time-correlated errors to improve the covariance realism efficiently. To estimate the parameters that govern the stochastic noise, the SCP model is combined with a previously developed uncertainty quantification method. Such method receives as input estimated and propagated orbits, quantifying the uncertainty of the system offline of the orbit determination and propagation processes, not requiring modification of operational Space Surveillance and Tracking (SST) systems. In this work, real radar observations from the Spanish Space Surveillance Radar (S3TSR) are used to test the covariance realism improvement of the SCP method for several RSOs at different altitudes with respect to deterministic constant error models. The results analyse the physical interpretation of the estimated noise parameters with real data, while also evaluating key metrics to assess covariance realism such as Cramer-von-Mises metric and covariance containment.