Vision-based Deck Estimation for Autonomous Ship-board Landing

Open Access
- Author:
- Truskin, Benjamin Louis
- Graduate Program:
- Aerospace Engineering
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- April 26, 2013
- Committee Members:
- Jacob Willem Langelaan, Thesis Advisor/Co-Advisor
- Keywords:
- Vision
Estimation
Aerospace
Engineering - Abstract:
- Piloted landings on ship decks are a daunting challenge where life critical missions require operations to continue even under adverse conditions. Automating the landing process for a helicopter on any landing site would create a safer environment in many applications. The research detailed in this thesis was motivated by an even more difficult and specific situation: attempted landing of a small, payload limited UAV on a moving, non-cooperative or mildly cooperative ship deck. While current capabilities do include landing an autonomous rotorcraft onto a moving deck, this can only be performed under very restrictive conditions involving a cooperative ship deck and calm seas. The goal for this system is to have the generality necessary to successfully operate in a variety of sea conditions, and with many potential ship types. This thesis details the development and simulation demonstration of a state estimation system that only requires the use of a monocular camera, an inertial measurement unit, and GPS. The INS/GPS is required to determine vehicle states. The camera provides information about landmark bearings relative to the UAV, but nothing else. By combining these two sensor systems, it is possible to estimate deck state, thus enabling the ability to autonomously land. This thesis presents Unscented Kalman Filter based implementation that utilizes a generic second order kinematic model driven by zero mean Gaussian noise for the ship deck motion model. Use of this generic model contains unmodeled dynamics, but is not particular to any ship. The goal is to estimate position, attitude, velocity, and angular velocity for the deck. Estimator performance is reviewed using single run and multiple Monte Carlo simulations where the deck is subjected to a variety of wave conditions and various wave modeling techniques.