This paper studies the localization problem of Wireless Sensor Network (WSN) in Indoor Non Line of Sight (NLOS) environment. Under condition of the optimization estimation framework, an error function to be optimized is firstly proposed, and then based on advanced Multidimensional Scaling-MAP (MDS-MAP) method, the coupling relationship among the unknown variables is analyzed, and finally an optimization localization algorithm based on improved MDS-MAP is designed. Without any prior environmental knowledge, the proposed algorithm can be conducted merely with distance information suitable for WSN in unknown and complex environment. Simultaneously, the algorithm does not need any redundant information (e.g. angle and signal strength), thus possessing the properties of low hardware requirement and low cost. The effectiveness of the proposed algorithm are illustrated through several simulations examples. And the clustering analysis method is adopted to analyze the localization precision with different numbers of basis stations. According to the simulation results, the approximate optimal localization results with low cost are reached in 3-dimensional space when the number of the basis stations is six.