THE SIMPLEX ALGORITHM FOR . UNDERWATER SOURCE LOCALIZATION
The objective of this thesis is to find an effective method for underwater acoustic source localization in a multipath deep-ocean environment. The model discussed in this thesis consists of a submerged three-sensor receiver system and a broad-band noise source. Only two propagation paths (direct path .and surface reflection path) from the source to the receiver array are considered, and the time difference of arrival (TDOA) between every two paths are used as the raw data for source localization. After a brief description of the multipath environment and the system geometry, three analytical methods are discussed in this thesis: linear approximation method, spherical interpolation (SI) method and hybrid method. These methods can give a quasi-optimum least-squares estimate of the source location with a relatively low computational cost. The accuracy of all these analytical methods decreases rapidly with the increasing source-receiver distance. This limits the application of these methods in a very small area around the receiver array. A numerical method, called the simplex method, is then discussed. As an iterative algorithm, the simplex method can get an excellent accuracy of localization with a relatively high computational cost. Moreover, as a numerical algorithm, the simplex method can use the absolute function (Ll norm) as error function, so as to make the estimator more robust than the least-squares estimator, especially when.there are "bad points" in the raw data. To reduce the computational time of the simplex method, the estimation result of the analytical method can be used as the initial simplex. The simulation results prove that this improvement can greatly reduce the iterative time of the simplex algorithm and make the algorithm more efficient. The simulation results in this thesis show that the performance of the simplex method is much better than that of the analytical methods. Therefore the conclusion of this thesis is that the simplex method can be used as an effective algorithm in the area of passive localization.