Decision Tree vs K-Nearest Neighbors: Machine Learning Based Wind Estimation for Unmanned Aerial Vehicles

In this work, we investigated the efficacy of two data-driven approaches towards wind estimation for a multicopter that follows a circular orbit rather than in a hover flight. The position and attitude of the multicopter were collected from flight simulations for a vareity of wind speeds surrounding the multicopter. Two regression methods based on the K-Nearest Neighbors (KNN) and Decision Tree (DT), respectively, were developed to produce wind estimates. The results show that both proposed methods are able to produce wind estimates with more than 98% accuracy. The DT-based method produces more accurate estimates than the KNN-based method while requires slightly more computational efforts.