Optimal Guidance for Ship Landing Using Monocular Vision

Small unmanned aircraft systems often rely on monocular cameras. Utilizing structure from motion (SfM) techniques monocular vision based systems are capable of estimating the position of an observed target. The estimation uncertainty depends on the relative motion of the camera and the target. This work presents an optimal control based landing guidance method that uses the Fisher information matrix (FIM) to increase the target observability along the approach. The target is parameterized using the Inverse Depth Parameterization (IDP) to reduce errors introduced by linearizing the measurement model during the propagation of the Fisher Information Matrix (FIM). By using Bernstein polynomials to approximate the optimal solution it is possible to reduce the computational complexity and impose constraints along the trajectory to ensure feasibility. The proposed concept is evaluated in numerical examples, demonstrating increased parallax in the observability-optimized trajectories and compliance with dynamic feasibility requirements.