A GPR Based Novel Approach for Aerodynamic Parameter Estimation from Flight Data

摘要

In this paper, a novel method based on Gaussian process regression(GPR) is proposed for the aerodynamic parameter estimation from the flight data. The new method GPR-Delta is an extension of the feed-forward neural network (FFNN) based Delta method. The GPR-Delta augments the philosophies of kernel-based nonparametric probabilistic models in the Delta method by replacing the FFNN. Efficacy of the proposed algorithm is examined by estimating the aerodynamic parameters using flight test data of two different categories of the aircraft. GPR-Delta estimated parameters are compared with the wind tunnel, Delta and Filter error method estimated aerodynamic parameters. Comparison results establish the GPR-Delta as a viable alternative method to this problem.