System identification of a small scaled helicopter using simulated annealing algorithm

Developing an autonomous helicopter requires designing of precise controllers, which can be a daunting task if system dynamics are not known accurately. Thus very accurate system dynamics should be available to design controllers. The main idea of this paper is to present Simulated Annealing (SA) algorithm as a tool in time domain parametric system identification of a RC helicopter (Align T-Rex 550L). In addition, Prediction Error Minimization (PEM) and Genetic Algorithm (GA) have been taken as reference identification algorithms for the purpose of comparison. The work includes collecting flight data and pre-processing of the recorded data, and time domain parametric identification of state space system for hovering condition. The rigid body dynamics of the helicopter is represented in the state space form that has 40 parameters. The accuracy of the identified system is verified by comparing estimated and actual responses, by Pearson Correlation Coefficient, also 90% confidence interval is calculated for each of the identified parameters. Results show a high level of correlation of the actual responses and estimated responses of the system identified using SA, because of its ability to jump out of local optima.