The propagation model based on experience has multi-valued problems in the practical application, which causes the RSSI and the distance not one- to-one mapping. This paper proposes an improved MDS-MAP localization algorithm, which uses propagation model based on distance interval ranging probability to eliminate the influence of multi-valued problems to ranging accuracy of the model, thereby improving measurement accuracy of ranging model. The simulation results show that the improved algorithm has better positioning accuracy when compared with the traditional MDS - MAP localization algorithm.