Nonlinear Model Predictive Path Following Control of a Fixed-Wing Single-Line Kite: An Experimental Study

Petr Listov

École polytechnique fédérale de Lausanne

Tuesday, July 25, 2017, 11:00

"Room 01-012, Georges-Köhler-Allee 102, Freiburg 79110, Germany"

A reliable flight control of Airborne Wind Energy (AWE) kites is known to be a challenging problem both from academic and industrial perspectives. In the recent years, several modeling and trajectory tracking strategies were suggested for the fixed-wing single-line class of energy kites. Validation and experimental results, however, for these methods are not available. This work aims at further studying AWE systems dynamics and application of optimization-based algorithms for path following control of energy kites.

In this talk a modelling, estimation and control framework for a fixed-wing single-line kite will be presented, as well as design of micro-scale AWE system prototype. The software stack consists of the kite simulation module, generic Extended Kalman Filter (EKF) implementation, and Nonlinear Model Predictive Path Following (NMPF) control algorithm. For the latter, we consider the constrained output path following problem of a closed path defined as a parametric curve in the output space of the system. The path parameter is treated as a virtual state, governed by an additional ODE. The Chebyshev pseudospectral collocation technique is chosen for the trajectory discretization to account for the inherently unstable and highly nonlinear nature of the system.

The indoor test platform comprises a radio controlled commercial propelled airplane, tailored with a nylon thread to the ground station unit. The station is capable of measuring line angles, tether force and control reeling in of the tether. The flying facility is equipped with OptiTrack motion capture system that provides ground-truth position and attitude measurements.

Throughout piloted flight experiments we identified viscoelastic parameters of the tether and adjusted values of some stability and control derivatives. Then the developed estimation and control framework is succesfuly tested to track a circular path in the flight simulator.