Abstract: Rank sequence-based localization method is a novel and high-accuracy wireless sensor networks (WSN) localization technique. The localization space is divided into distinct sub-regions and each is uniquely identified by a rank sequence. However, the localization errors for nodes on the edge of a region are rather large and they are not optimal in view of average localization errors. This paper proposes a new N-best rank sequence localization method. The best value N is first achieved using the random sampling for reference nodes based on a wireless channel fading model, and the coordinates for the target are then computed through selecting the top N rank sequences. We have conducted the simulation with 100 nodes, the outdoor experiment with 15 ZigBee physical nodes, and the air-raid shelter tunnel test with 10 ZigBee nodes. All the results have shown that our method reduces the average localization errors and improves the localization accuracy for nodes on the edge of the region.