Research project: Multiple-Source Localization Problems
The source localization problem in its general form is a nonconvex optimization problem, which typically involved rank constraints.
The source localization problem in its general form is a nonconvex optimization problem, which typically involved rank constraints.
However, when there is only one unknown source, it is known that the problem can be equivalently formulated as the generalized trust-region subproblem (GTRS) and hence can be solved to its global optimality. The methodology of GTRS breaks down when there are multiple unknown sources as there would be multiple quadratics constraints in GTRS. This project aims to establish a global theory as well as fast algorithms for the multiple-source case. It proposes to use the classical Euclidean distance geometry to formulate the problem as a distance problem, which has a close link to quadratic semi-definite programming and the problem of wireless sensor network localization. Real applications will also be looked into.