An EKF formulation for pose estimation of a landing platform fixed to a moving vehicle for an autonomous landing system

Resumo

Automatic landing systems are heavily dependent on sensors to map the environment around the aircraft
in order to support the subsequent control strategy to accomplish the task. This work proposes a discrete-time
Extended Kalman Filter (EKF) formulation, using camera measurements, to estimate the position and orientation
(yaw) states of a landing platform, fixed on top of a Ground Vehicle (GV). GV moves freely at constant speed
The camera is attached to an Unmanned Aerial Vehicle (UAV) and is pointed down. For each image taken by the
camera, a computer vision algorithm returns the relative position and attitude (pose) of each identified marker. Its
is adopted the open-source library Aruco for generating the printable square-based fiducial markers. Its adoption
is justified because it allows quick fixation to the landing platform, being easily detectable and providing a robust
determination of the relative pose. The EKF is formulated as a constant velocity model for the pose estimation.
The simulation consists of the UAV following the GV along a path, where the desired states are estimated from
noise-corrupted measurements. The proposition is validated via Monte Carlo simulation. The results showed that
the proposed formulation for the EKF is able to estimate the desired states when operating at low speeds.