Comparative Analysis of Innovation-Based Adaptive Kalman Filters Applied to AUVs Navigation*

The navigation of Autonomous Underwater Vehicles (AUV) is affected by some factors such as unknown noises degrading the position estimation, the impossibility to use high frequency signals in underwater and the instability of the water environment, accentuating the vehicles position errors when a conventional Kalman filter algorithm is used. The conventional Kalman filter assumes that the process and measurement noise matrices are constant, however, studies on adaptive Kalman filters have emerged promising to ensure stability by acting directly on these matrices. The Innovationbased Adaptive Estimation Adaptive Kalman Filter (IAE-AKF), uses the innovation history to ensure optimal filtering. A disadvantage of this method is that the presence of outliers in the sensors that accentuates the measurement noise and degrades performance of this filter. To mitigate outliers, Robust IAE-AKF (RIAE-AKF) evaluates the innovation sequence and corrects for possible anomalies that may degrade the filter estimates. This paper performs a comparative analysis between the conventional method (KF) and the adaptive methods (IAEAKF and RIAE-AKF) in AUV navigation applying the Monte Carlo method on simulated data for a sensory fusion navigation using an Inertial Measurement Unit (IMU), a Doppler Velocity Log (DVL) and a Pressure Sensor (PS)