In the mobile wireless sensor network(MWSN) of underguound coal mine, as the mobile node of the network, the determination of the spatial position of the mobile robot plays an important role in the normal operation of the network. Aiming at the problem that the accuracy and complexity of the three-dimensional positioning algorithm in wireless sensor networks are difficult to balance, this paper extends a probabilistic centroid localization algorithm based on received signal strength indication (RSSI) ranging in a two-dimensional plane to a long and narrow space in coal mines for three-dimensional positioning application of mobile robot nodes. The algorithm adopted the overlapping area,which should be locating spherical shell of anchor node under a certain outstanding degree, replacing the whole distribution area of the robot node estimated position.Then the probability-density-centroid of overlapping area was figured out as the estimated value of the robot node. The position algorithm have been verified by a simulation experiment. The result manifest that the Positioning accuracy can reach 0.13 in long and narrow space.