A novel intelligent adaptive Kalman Filter for estimating the Submarine's velocity: With experimental evaluation

In this paper, an intelligent adaptive algorithm based on the integrated Inertial Navigation System (INS) is presented for estimating the velocity of an Autonomous Underwater Vehicle (AUV). The most common algorithm for incorporating the navigation data is Kalman Filter (KF). In the conventional KF algorithm, the covariance matrix of measurement noise is considered as a constant through running time. The noise covariance in the Adaptive Kalman Filter (AKF), is estimated through processing the innovation sequence inside a window of constant length. In actual conditions, the correct window length to achieve the best estimation, depends on the operational conditions. In this paper, proper window length in the adaptive algorithm is estimated by processing the probability density function of the innovation sequence at each time step; so the window length in the proposed algorithm is variable due to condition changes. The proposed integrated navigation system consists of a three-axis Inertial Measurement Unit (IMU) and a three-axis Doppler Velocity Log (DVL). The performance of the proposed system is evaluated through four sea tests using an AUV. Experimental results show that the proposed system has superior performance than the conventional KF algorithm and is similar to the optimum AKF (AKF with the best window length).