Predicting the navigation performance of underwater vehicles

In this paper we present a general framework for predicting the positioning uncertainty of underwater vehicles. We apply this framework to common examples from marine robotics: standalone long baseline (LBL) positioning and integrated LBL reference and Doppler velocity log (DVL) dead-reckoning. The approach is based on formulating positioning as an estimation problem. Using simple sensor models for the most common information sources, we show how the the Cramer Rao lower bound can be used to predict the system-level navigation performance. The resulting three dimensional covariance matrix is then summarized using scalar performance metrics based on the notion of dilution of precision (DOP), a well-known concept from the global positioning system (GPS) community. To illustrate this general tool, we present the answers to a few particular questions: · How does the baseline length affect the solution for standalone LBL positioning? · When using DVL and heading odometry, how precise is the combined DVL/LBL solution? · For an integrated DVL/LBL solution, what is the required update rate from the absolute reference to constrain odometry drift? · What is the relative importance of heading, odometry and range precision for overall performance? We substantiate our estimation framework through experimental evidence which shows that the analytical predictions are consistent with performance in the field.