Adaptive Wind Field Estimation Using an Empirical Bayesian Approach

Abstract

In lidar-based gust load alleviation, the wind profile ahead of the aircraft cannot be measured directly but has to be reconstructed (estimated) based on the acquired line-of-sight measurements. Such wind reconstruction algorithms typically include regularization in order to adequately handle the noise within the data. This paper presents an empirical Bayesian approach to choosing optimal regularization parameters for any given set of measurements. Using simulations of flight through turbulence, the Bayesian approach is compared with a former approach (based on engineering guess) and an omniscient optimizer, which yields the best achievable results for a constant set of parameters by using the full knowledge of the wind field. The Bayesian approach is shown to outperform the engineering guess and performs close to the omniscient optimizer while purely relying on the lidar measurement data.