A novel indoor localization algorithm was presented

which employs robust estimator to identify and restrain ranging outliers or gross errors and uses DFP （davidon fletcher powell） method to majorize the global object function with a convergence within 2 steps. It first divides all the ranging measurements into three different domains （effective in- formation

usable information and bad information） according to the corresponding residual errors

and then adopts dif- ferent weighting scheme （maintaining

down-weighting

rejecting） through self-adaptation during iterative process. Ex- tensive simulation results confirm that this proposed localization scheme outperforms remarkably traditional least squares （LS）

which do not employ outlier identification and restraint.