Performance of GPS and IMU sensor fusion using unscented Kalman filter for precise i-Boat navigation in infinite wide waters

The Unmanned Surface Vehicle (USV) navigation system needs an accurate, firm, and reliable performance to avoid obstacles, as well as carry out automatic movements during missions. The Global Positioning System (GPS) is often used in these systems to provide absolute position information. However, the GPS measurements are affected by external conditions such as atmospheric bias and multipath effects. This leads to the inability of the stand-alone GPS to provide accurate positioning for the USV systems. One of the solutions to correct the errors of this sensor is by conducting GPS and Inertial Measurement Unit (IMU) fusion. The IMU sensor is complementary to the GPS and not affected by external conditions. However, it accumulates noise as time elapses. Therefore, this study aims to determine the fusion of the GPS and IMU sensors for the i-Boat navigation system, which is a USV developed by Institut Teknologi Sepuluh Nopember (ITS) Surabaya. Using the Unscented Kalman filter (UKF), sensor fusion was carried out based on the state equation defined by the dynamic and kinematic mathematical model of ship motion in 6 degrees of freedom. Then the performance of this model was tested through several simulations using different combinations of attitude measurement data. Two scenarios were conducted in the simulations: attitude measurement inclusion and exclusion (Scenarios Ⅰ and Ⅱ, respectively). The results showed that the position estimation in Scenario Ⅱ was better than in Scenario Ⅰ, with the Root Mean Square Error (RMSE) value of 0.062 m. Further simulations showed that the presence of attitude measurement data caused a decrease in the fusion accuracy. The UKF simulation with eight measurement parameters (Scenarios A, B and C) and seven measurement parameters (Scenarios D, E and F), as well as analytical attitude movement, indicated that yaw data had the largest noise accumulation compared to roll and pitch.