Automatic testbed with a visual motion tracking system for airborne wind energy applications

Abstract

Abstract The architecture and a flight test campaign of a small‐scale testbed aimed at aerodynamic and dynamic characterization of airborne wind energy systems are presented. The testbed involves a two‐line rigid‐framed delta kite and an automatic ground station for the lateral control of the kite and reel‐in/reel‐out of the two tethers. The environment, and the states of the kite, the tethers and the actuators are measured by a set of on‐ground and onboard sensors that include, among others, an inertial measurement unit, GNSS receivers, load cells, actuator encoders, a wind station, and a visual motion tracking (VMT) system based on three cameras and an artificial neural network (YOLOv2). The results of a 5‐min flight, including the take‐off, cross‐wind flight, and landing, were used to analyze the capabilities of the testbed. It was shown that the time derivative of the kite course angle exhibits a linear correlation with both the delayed steering input and the delayed differential tether tension, being the dispersion lower for the latter. The intrinsic and extrinsic calibrations proposed for the VMT system led to a good agreement between the estimation of the kite position and course angle provided by the VMT system and the onboard computer. Moreover, although the YOLOv2 algorithm failed in the detection of the kite within around 5% of the images, the simultaneous non‐detection from the three cameras was below 0.1% during the full flight. Such a reliability suggests that a VMT system can be used as a redundant or backup sensor for the GNSS.

Keywords

• artificial neural network
• Kalman filter
• open‐loop control
• rigid‐framed delta kite
• visual motion tracking