Extended Linear Regression and Interior Point Optimization for Identification of Model Parameters of Fixed Wing UAVs

Abstract

The paper deals with the identification of the system parameters in the nonlinear dynamic model of fixed wing UAVs. Fixed wings airplanes are popular in long distance applications and have to be modelled accurately to guarantee efficient control properties. Different methods are suggested to solve the parameter estimation beginning with the standard linear regression (LR) and continued with its extension (ELR), the optimization using interior point methods and finally using the FireFly technique, which is a metaheuristic algorithm. The methods illustrate the convergence and the speed of these approaches. The known default parrameter values of a Sekwa UAV were used to demonstrate that the elaborated identification methods can also reconstruct the numerical values of the dimensionless system parameters embedded into the nonlinear model using the physical weighting functions. After the extension of linear regression, MinMax optimization algorithm was used to get the best and optimal solution and reconstruct the parameters of the Sekwa aircraft. FireFly optimization gives also comparable results with minmax method. Flight data was needed for simulation and testing the different approaches. Matlab and toolboxes were used as simulation software. The results showed the estimated parameters are more accurate than linear regression estimation. Even though it is a small improvment this will reflect in all calculations.