Numerical Techniques for Efficient Sonar Bearing and Range Searching in the Near Field Using Genetic Algorithms

Summary

This article describes a numerical method that may be used to efficiently locate and track underwater sonar targets in the near-field, with both bearing and range estimation, for the case of very large passive arrays. The approach used has no requirement for a priori knowledge about the source and uses only limited information about the receiver array shape. The role of sensor position uncertainty and the consequence of targets always being in the near-field are analysed and the problems associated with the manipulation of large matrices inherent in conventional eigenvalue type algorithms noted. A simpler numerical approach is then presented which reduces the problem to that of search optimization. When using this method the location of a target corresponds to finding the position of the maximum weighted sum of the output from all sensors. Since this search procedure can be dealt with using modern stochastic optimization methods, such as the genetic algorithm, the operational requirement that an acceptable accuracy be achieved in real time can usually be met.

The array studied here consists of 225 elements positioned along a flexible cable towed behind a ship with 3.4m between sensors, giving an effective aperture of 761.6m. For such a long array, the far field assumption used in most beam-forming algorithms is no longer appropriate. The waves emitted by the targets then have to be considered as curved rather than plane. It is shown that, for simulated data, if no significant noise occurs in the same frequency band as the target signal, then bearing and range can be estimated with negligible error. When background noise is present (at -14dB), the target can normally still be located to within 1% in bearing and to around 5% error in range. Array shape uncertainty worsens this performance but it is shown that it is still possible to accurately determine the target bearing and to provide approximate estimates of range in such circumstances. Finally, the ability of the approach to track the sources over a period of time and with evolving array shape is demonstrated.