The navigation system of an autonomous underwater vehicle for Antarctic exploration

Abstract

In the context of the Italian Scientific Program for Antarctica an underwater vehicle is being developed that will be capable of operating autonomously for many hours under the Antarctic shelf or in sea areas covered by floating ice blocks. The vehicle must therefore carry a navigation system satisfying severe constraints on accuracy, size, and power consumption. During navigation the vehicle cannot use low-cost systems such as magnetic compasses (ineffective near the Earth poles), GPSs (which cannot be used underwater), or acoustic transponders (because of the long range of the missions). To estimate its own heading and position, the vehicle can only exploit measurements of acceleration, rotation rate, and velocity. The optimal integration of inertial and velocity measurements is accomplished by using a Kalman filter for correcting the effect of the biases of the inertial sensors (which would cause relevant long-term position errors) by exploiting a velocity measurement. Conceptually, the filter blends the high-frequency components of the inertial estimation with the low-frequency components of the velocity measurement, giving a long-term position estimation that is much more accurate than that obtainable using only accelerometers. A Doppler-dumped inertial navigation system has been developed using the recently available subsystems and implementing innovative signal processing techniques, obtaining a satisfactory trade-off between reliability, cost, accuracy, size, and power requirements. The inertial sensors (three accelerometers and three ring laser gyroscopes) are packaged in a small-size and relatively low cost, strap-down inertial unit. The velocimeter is a compact low-power Doppler sonar specifically designed for underwater vehicles. Signal processing is implemented on a general purpose computer having small space and power requirements. The performance of the Kalman filter was optimized by using the "indexing" technique, which consists of periodically rotating the inertial unit to increase the observability of the Kalman filter states and thus improve the correction of the biases of the inertial measurements. In the paper, the navigation system is described and its performance analyzed, showing the effectiveness of the indexing technique in improving the Kalman filter performance both in convergence speed and in estimation accuracy.