Matched-field inversion techniques are widely used for source localization and geoacoustic parameter estimation. These inversion methods correlate the received data with modeled data and find the model parameters which provide the maximum correlation. However, when a large number of unknown parameters is involved, many modeled data need to be generated and correlated with the observed data and thus, matched-field inversion can be computationally intensive. An optimization process applied to matched-field inversion is often required to accelerate the inversion process.
In this work, tabu is applied to matched-field inversion for source localization and environmental parameter estimation. Tabu is a global optimization technique which proceeds by finding the best model in a local neighborhood, where a best model is defined as the set of parameter values that provides the maximum correlation in a given neighborhood. However, the search moves beyond local areas by maintaining records of past moves. Using historical information, the approach avoids certain paths. Thus, tabu limits the search space and redefines neighborhoods in each iteration. Tabu is evaluated through a comparison to fast simulated annealing.
To improve efficiency, a tabu approach is also developed for parameter estimation in a rotated coordinate system. Rotation is achieved through the identification of combinations of parameters that affect acoustic field computations.