Abstract: Localization is an intrinsic issue in wireless sensor networks. To make up for the deficiencies of the classic multi-dimensional scaling algorithm MDS-MAP as to its localizing accuracy and complexity in matrix computation, the RSSI-GA algorithm was proposed. In its multi-dimensional scaling process a dissimilarity matrix is directly formed according to the value of wireless signal intensity. Analyzing the dissimilarity of individuals and geometric constraint relationship the distances between nodes, a mathematical model was developed with the positions of unknown nodes as parameters. Then, the genetic algorithm was introduced to solve the model so as to compute the node coordinates directly. Simulation results showed that the RSSI-GA algorithm can reduce greatly the cost for computation and improve the accuracy of localization efficiently.