Abstract: Many of location algorithms for sensor networks are only suitable for relatively uniformly de- ployed networks. Consequently, their location precision is poor if the sensor networks are depolyed nonu- niformly. However, the sensor networks are often deplayed nonuniformly in practice because of the moni- tored regional terrain or application requirements. The virtual hole is introduced in order to optimize the distance between nodes. Based on the multidimensional scaling technology, a novel location algorithm, called Multi-Dimensional Scaling via Distance-Optimized (MDS-DO), is proposed for nonuniformly de- ployed sensor networks. Its location performance is simulated and analyzed in semi-C network. The re- sults shows that the performance of MDS-DO is much better than that of the classical MDS-MAP in the case of different node quantities and connectivity. Furthermore, the average location error of MDS-DO is decreased about 90% compared with MDS-MAP in the case of high connectivity.

Key words: information processing, wireless sensor network, nonuniform deployment, location, multi- dimensional scaling