System Identification for Yaw Dynamics Prediction Model of an Autonomous Underwater Glider: Comparison of N4SID and ARX Models Using Preprocessed Data

This paper aims to compare two algorithms of system identification between Auto-Regressive with eXogenous inputs (ARX) and Numerical Subspace State Space System Identification (N4SID). This comparison is based on the development of a Model Predictive Control (MPC) for heading control of an Autonomous Underwater Glider (AUG). The comparison was done by compared best-fit values of those models, made from data through a noise filtering process. The filtering process is using a Maximum Likelihood-based Kalman Filter (ML-based KF). ML is used to get the optimal parameters of Q and R value for K <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$F$</tex>, thus that the optimal value for both parameters can be achieved. Both models have good accuracy performance on best-fit percentage and have low errors. Nevertheless, the ARX model demonstrates a bit advantage over N4SID, achieving the highest best-fit percentage than N4SID model. The ARX achieved the best fit percentage of 71.76<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> when the na and nb tuned at 6 and 4, respectively with the errors achieved 3.127 for MSE and 1.768 for RMSE. Then the N4SID achieved the best fit percentage of 71.75% at the seventh-order tuning with MSE and RMSE are 3.1586 and 1.7725, respectively. This algorithm will be used as a prediction model for MPC, in future.