Wind Estimation Using Quadcopter Motion: A Machine Learning Approach

摘要: In this article, we study the well known problem of wind estimation in atmospheric turbulence using small unmanned aerial systems (sUAS). We present a machine learning approach to velocity based on quadcopter state measurements without sensor. accomplish by training long short-term memory (LSTM) neural network (NN) roll and pitch angles position inputs with forcing velocities as targets. The datasets are generated simulated turbulent fields. trained is deployed estimate winds Dryden gust model realistic large eddy simulation (LES) near-neutral boundary layer (ABL) over flat terrain. resulting NN predictions compared triangle that uses tilt angle an approximation airspeed. Results from indicate LSTM-NN predicts lower errors both mean variance local field approach. work reported article demonstrates potential for sensor-less has strong implications large-scale low-altitude sensing sUAS environmental autonomous navigation applications.