Information-Aware Model Predictive Control for Satellite Inspection

Abstract

Autonomous spacecraft inspection requires trajectories that satisfy safety and control constraints while enabling the collection of informative measurements about a target spacecraft. Traditional guidance and control methods typically decouple estimation from control, resulting in trajectories that do not explicitly optimize sensing geometry. This work presents a model predictive control (MPC) framework that incorporates estimation covariance in the control objective using a formulation inspired by dual control and covariance steering. The estimation covariance evolves according to a linear Kalman filter, and the measurement model depends on the relative geometry between the agent spacecraft and the target. By embedding the covariance dynamics within the MPC problem, the resulting trajectories account for measurement quality, actively reduce uncertainty, and improve observability in the estimated features of the target. The problem is formulated using relative motion dynamics via the Hill-Clohessy-Wiltshire equations with constraints on control input, relative distance, and terminal maximum covariance. Numerical simulations demonstrate that the proposed framework generates feasible inspection trajectories that actively reduce estimation covariance of points of interest on a target while satisfying input and safety constraints of the agent. A mesh analysis of initial conditions further illustrates how feasibility and the value function of the trajectory depend on the initial conditions and constraint activity.