GPS localization problem: a new model and its global optimization

We establish a new fractional squared least squares (FSLS) optimization model for the GPS localization problem. It provides more accurate solutions than the classical squared least squares model. We reformulate (FSLS) as a univariate optimization, where the functional evaluation corresponds to the generalized trust region subproblem. We employ the branch and bound algorithm to globally solve (FSLS) and establish the convergence. It further motivates a much faster iterative heuristic algorithm. Numerical examples are presented to show the accuracy of the new model (FSLS) and the efficiency of the two algorithms.