Autonomous MAV Landing on a Moving Platform with Estimation of Unknown Turbulent Wind Conditions

Abstract

This paper presents an autonomous landing of a micro aerial vehicle (MAV) on a moving platform immersed in turbulent wind conditions. We estimate the 3D wind vector acting on the vehicle using a model-based and a deep learning-based approach. A disturbance-aware boundary layer sliding controller then uses this estimation to generate a control input that provides trajectory tracking guarantees in the presence of unknown, but bounded disturbances. The approach presented integrates our previous works on control and estimation, and we show its performance in a challenging setting. The experiments show that our methods enable fast landing on a moving platform in turbulent, unknown wind conditions.

Experimental verification of an LiDAR based Gust Rejection System for a Quadrotor UAV

“…The computation load raised by onboard wind estimation is another limitation for smaller drones. Wind estimation can also be achieved by using onboard wind measurement devices [14], [24]. While these methods can measure the wind at the exact location of the drone or provide the preview wind information for control, the measurement values are contaminated through the influence of the drone propellers on the surrounding air.…”