Quantcast

Vision-based Autonomous Landing Control for Unmanned Helicopters

Research paper by Panos Marantos, George C. Karras, Panagiotis Vlantis, Kostas J. Kyriakopoulos

Indexed on: 28 Oct '17Published on: 04 Oct '17Published in: Journal of Intelligent & Robotic Systems



Abstract

This paper presents the design of a complete control system for the autonomous landing of unmanned flybarless helicopters on known stationary visual landmarks. A state estimator based on the complementary filters notion, estimates the position, translational velocity and attitude vectors of the vehicle by fusing data acquired from the on–board camera and an Inertial Measurement Unit. A vision-aided nonlinear model predictive controller is designed for the landing motion of the helicopter, assuming that the on–board camera is rigidly (i.e., no additional Degrees of Freedom (DOF)) attached on the vehicle. Although the under–actuated character of the helicopter dynamics introduces counter–goals for minimizing the error between the vehicle and the landmark, the proposed control scheme guarantees, via hard nonlinear constraints, that the landmark will always be kept inside the camera field of view during the landing procedure. In order to simplify the derived algorithm without violating the robustness of the proposed controller, we reformulate the translational helicopter dynamics in order to reduce the number of the unknown model parameters to a minimum. Moreover, a parameter/disturbance observer is designed for estimating simultaneously the vehicle’s unknown dynamic parameters as well as the induced disturbances. The efficacy of the proposed landing scheme is evaluated via a set of experimental and simulation results, using a small–scale flybarless helicopter.