Due to continuous and unplanned urbanization, biases and probability of occurrences of non-line-of-sight (NLOS) errors can be drastically enlarged in macro-cellular smart urban environments. This paper presents a new robust estimation approach to mobile tracking improvements based on adequately tackling NLOS errors. To cover the dynamics of a mobile station, we cast a wireless localization problem into a Markov state transitioned system framework. We then introduce a screening scheme for reducing NLOS effects on received measurements, and a M-estimator. These enable us to develop an enhanced robust-regression-based interacting multiple-model (enhanced R-IMM) algorithm. Simulations demonstrate the advantages and superior performance of the enhanced R-IMM.