Ship Deck Tracking in High Sea States via Distributed Estimation

Safe and reliable landing of autonomous UAS on moving maritime platforms poses a significant challenge. This paper presents a decentralized and distributed extended information filter for ship deck state estimation and landing by cooperative UAS. Each UAS estimates the relative pose of the ship using fiducial markers and computer vision, combined with a second-order kinematic model of the ship, fused in the information filter. Algorithm performance and functionality are evaluated in a simulation environment that models different sea states using a two-dimensional directional Pierson-Moskowitz ocean spectrum and a ship dynamic model (so that dynamics that are not modeled in the information filter are in the simulated ship model). The simulation tests several deck state scenarios, including cooperative and evasive ship behavior. The filter is initialized based on anticipated sea conditions and the assumed cooperativeness of the ship during the mission. The algorithm leads to accurate deck state estimates both under full-rate communication and during intermittent communication losses among UAS.