Development of error reduction model using Bayesian filter for AUV navigating under moving ice

Doppler Velocity Logs (DVL) can provide a simple under water navigation aid for Autonomous Underwater Vehicles (AUV) by measuring relative velocities with respect to the speed over ground. A valid reference velocity is difficult to calculate when this approach is applied under a moving frame of reference such as drifting ice. The primary challenge of under-ice localization is to accurately estimate the AUV location and its trajectory in the global coordinate system when DVL measurements are being made relative to a constantly drifting ice surface. In this paper, the author introduces and compares two types of error sources, scale factor error of DVL and navigation error due to ice drift. An error reduction model using a Bayesian filter algorithm is developed for improved estimations, in conjunction with a novel correction method for accurate AUV navigation under ice. The concept of shift factor is introduced in this paper as the key to solve both error sources. Having the knowledge of the true beacon location, shift factors in vector quantity are extracted based on the collected relative velocity profiles by DVL. The shift factors are directly applied to update the final AUV location. The result presents approximately 70.8% of maximum error reduction. The impact of survey pattern, bearing angle to the beacon, pinging frequency on the accuracy of the vehicle localisation are discussed.