Visual servoing MPC framework for shipboard autonomous landing

Abstract

Autonomous shipboard landing under harsh sea state conditions is a challenging task, with platform oscillations being one of the critical challenges to expand the flight envelope. Conventional GNC systems usually rely on deck-mounted equipment to supply relative pose data, which limits interoperability within a naval fleet. Image Based Vision Servoing (IBVS) controllers provide a complementary solution that only relies on feature detections and image plane processing. However, current approaches are limited to mild sea state conditions and mainly consider current ship states within the landing criteria. This can increase the risk of unsafe conditions under harsh sea states, such as rollover. This paper introduces a framework that integrates visual servoing with a nonlinear model predictive control (NMPC) to address the attitude oscillations of the landing pad by introducing a custom cost barrier function. Vessel states are forecasted in real-time using a Fast Fourier Transform (FFT) combined with a Discrete Kalman Filter (DKF), enabling short-term prediction without requiring prior knowledge of vessel dynamics. Simulation results and flight test validation demonstrate that the proposed approach achieves a higher success rate under high sea state condition compared to both conventional IBVS based on current or predicted states.