Improvement of Filter Estimates Based on Data from Unmanned Underwater Vehicle with Machine Learning

In this study, the mathematical model of unmanned underwater vehicle is obtained in 6 DOF. The navigation sensor data are generated from mathematical model response. Extended Kalman filter and Unscented Kalman filter is applied to estimate noisy sensor data. For the EKF, nonlinear model is linearized around the equilibrium points. For the UKF, nonlinear system model is used. The estimation performance of EKF and UKF are compared. Estimation has been improved by applying support vector machine algorithm, which is machine learning, for unscented Kalman filter estimates. All this study is modeled in MATLAB/Simulink and PYTHON environment.