Non-Acoustic Sensor Array Localization Using Evolutionary Programming

Accurate localization of an array of sensor nodes is often a problem in ocean surveillance areas where in-situ surveying is not possible. This results from the inability to completely control an array during deployment. Ocean currents, ship motion, and array line flexure often make post deployment calculation of final node positions largely a matter of guesswork. Different optimization techniques are researched as estimators for sensor node positions. In this study, the sensors consist of an array of total field magnetometers interleaved with vector magnetometers and electromagnetic sensors. A surface ship with known course and speed traveling near the array generates a magnetic signature which is picked up by the array. A far-field dipole model is assumed and the parameters are estimated by minimizing the least-squared error discrepancy between the model and the sampled time series. The robustness of optimization using evolutionary programming is demonstrated where traditional gradient descent and variable-metric optimizers often fail for this parameter estimation problem.