An observability-enhanced GNC architecture for non-cooperative space target interception using angles-only navigation

This paper addresses the active defense of high-value critical space assets, which possess limited maneuverability and cannot afford fuel-expensive collision avoidance maneuvers. Nanosats equipped with passive optical camera offer a low-cost, flexible, and wide-coverage solution for intercepting non-cooperative targets like space debris. However, the use of monocular vision which provides only angle measurements without depth information leads to inherently poor observability in angles-only relative navigation. To address this, this paper proposes a guidance, navigation, and control (GNC) architecture designed to enhance observability, ensuring precise interception of non-cooperative targets. Firstly, by employing a high-precision relative dynamic model accounting for perturbations as the foundation, the GNC architecture is designed. This high-fidelity model not only supports the guidance and control functions but also improves system observability, which is explicitly analyzed. Secondly, A square-root cubature Kalman filter (SRCKF), selected for its strong nonlinear estimation capability, is then employed to track the target’s position and velocity states. Subsequently, an interception control strategy is formulated using model predictive control (MPC). A dynamic observability constraint derived from the Fisher Information Matrix (FIM) is integrated into the MPC to ensure the control inputs actively satisfy the evolving observability requirements during the interception mission. Finally, numerical simulation results demonstrate that the proposed GNC architecture enables nanosats to accurately reach the designated interception point. Furthermore, the high-precision dynamic model and the integrated observability constraints collectively ensure sustained high observability while reducing fuel consumption, making this approach suitable for the protection of critical space assets.