Covariance Steering for Powered-Descent Guidance with Embedded Navigation and Tracking

This paper presents a multirate output-feedback chance-constrained guidance framework for powered-descent guidance under uncertainty, coupling a high-frequency estimator/tracker with a low-frequency covariance-steering planner. A gain-scheduled inner-loop linear-quadratic regulator stabilizes tracking about a reference trajectory using high-frequency extended Kalman filter estimates, while an outer-loop planner computes the feedforward thrust and a covariance-shaping feedback gain on a coarser grid to regulate dispersion and enforce probabilistic constraints. The closed-loop mean and covariance are propagated via a linear covariance recursion, enabling a convex surrogate formulation of the covariance-steering problem. To enhance computational efficiency, semidefinite constraints are approximated by second-order cone constraints, and the induced conservatism is quantified and compensated. Numerical simulations demonstrate closed-loop feasibility, probabilistic constraint satisfaction, and close agreement between predicted and Monte Carlo covariances, while also revealing performance tradeoffs associated with inner-loop tuning and conservatism calibration.