Abstract: In view of problem of big positioning error of underground personnel positioning method based on weighted centroid algorithm, a fusion algorithm of underground personnel positioning based on self-organizing competitive network was proposed. The algorithm uses learning screening capacity of self-organizing competitive network, and screens out actual RSSI value close to the theoretical value by group training, then finds valid coordinates for the weighted centroid algorithm, and calculates unknown node position based on weighted centroid algorithm. The Matlab simulation results show that the positioning accuracy of the algorithm is significantly higher than the original weighted centroid algorithm.