The development of an acoustic navigational aid for autonomous underwater vehicles

Boltryk, Peter Jan (2004) The development of an acoustic navigational aid for autonomous underwater vehicles. University of Southampton, School of Engineering Sciences, Doctoral Thesis.

Record type: Thesis (Doctoral)

Abstract

This thesis summarises the research contribution to an ongoing project that is concerned with the development of a correlation velocity log (CVL), aimed specifically at autonomous underwater vehicles. The main objective of this research has been to improve the performance of a prototype CVL instrument that has been developed through collaboration between Chelsea Technologies Group, H Scientific Ltd and the University of Southampton.
A CVL is an acoustic-based navigational aid that uses a transmitter and an array of receiving elements in order to estimate velocity through the water relative to acoustic scatterers. A pair of short-duration acoustic pings is projected vertically downwards towards the seabed. Similarities in the echoes detected across the receiving array can be used together with knowledge of the acoustical array spatial relationships to estimate velocity. Similarity between received signals is determined by calculating a correlation coefficient between all combinations of receivers. The spatial distance separating the pair of receivers exerting the highest correlation is used together with the interpulse interval in order to estimate velocity of the craft.
The first half of this research work has been concerned with analysing existing data collected during past trials of the prototype system; in particular, the effect of phase differences across the acoustic receiver array has been investigated in depth. Analysis of the phase showed that sampling delays within the processing system and a phase inversion of one receiver's signals were causing unexpected phase variations across the acoustic array. Left uncorrected, these phase errors were undoubtedly reducing the effectiveness of the peak finding routines in the processing of data. Compensation algorithms have been derived to correct the phase within the DSP.
The second major section of work has concentrated on investigating peak estimation techniques for peak finding on a correlation surface that is central to the operation of the CVL. The methods compared include the indices of the highest point in the dataset, fitting of an axisymmetric quadratic model using either least squares or a non-linear, iterative implementation of maximum likelihood estimation, radial basis functions and Gaussian Processes. Numerical simulations demonstrate that radial basis functions provide the most robust method for peak estimation in this context. However, for simplicity of implementation on the DSP system, least squares remains the favoured method, offering respectable peak estimation accuracy throughout the measurement area.

This record has no associated files available for download.