Visual servoing model predictive control for autonomous shipboard rotorcraft landing in high sea states

Abstract

Autonomous rotorcraft landing on ships under harsh sea state conditions is a challenging task. Wave-induced oscillations of the landing deck require precisely timed touchdowns to prevent unsafe conditions such as rollover. Moreover, a key challenge is to enable autonomous landing without relying on complex hardware installed on the deck. This paper presents a novel approach that integrates visual servoing with model predictive control to overcome the requirement for hardware on deck while introducing vessel motion into the landing decision using cost barrier functions. Short-term online forecasting of landing pad states is achieved using a combination of fast Fourier transform and Kalman filter. The proposed model was evaluated through scaled-down simulations and indoor flight tests replicating harsh sea conditions. Two alternatives were compared to the baseline image-based visual servoing: a forecast-enhanced IBVS and the proposed MPC. Simulations show that the baseline visual servoing often fails to meet roll constraints at touchdown since it relies on current vessel states. The proposed controller outperforms both the baseline and its forecast-enhanced version, achieving an improved landing success rate under Sea State 6 with significant roll oscillations. Success criteria included roll constraints and touchdown accuracy, supported by a statistical database with multiple landing trials.