Sensing Wind from Quadrotor Motion

Abstract

This paper presents a comparison of three model-based algorithms to estimate the wind vector from quadrotor motion, based on increasingly refined models. The three vehicle motion models are the kinematic particle model, dynamic particle model, and rigid-body model. These motion models were characterized using results from wind-tunnel experiments and flight tests. The frequency response characteristics of each model were then examined to determine the range of wind fluctuations captured with each model. Analysis shows that the bandwidth of each wind-sensing algorithm increases with the fidelity of the model being used. To validate wind estimation results based on the three motion models, field experiments were performed in which the quadrotor was stationed in hover beside a sonic anemometer mounted 10 m above the ground on a weather tower in Innisfree, Virginia. Analysis of the resulting data indicates that the accuracy of wind direction estimates was comparable for all three model-based estimation schemes, but the rigid-body model yields more accurate estimates of wind speed. Because the vehicle dynamics will vary with small changes to the platform or operation at different altitudes, the relative ease with which a rigid-body model may be identified further supports its use for wind estimation from quadrotor motion.