Compared with the traditional position technology, the method based on Received Signal Strength (RSS) has lower requirements and operation complexity for hardware equipment, which is suitable for underwater sensors position. Underwater acoustic source sensor nodes are randomly distributed, and the number is limited in a large area in water, so the underwater acoustic sensor position network is sparse. In view of this characteristic, the Compressed Sensing (CS) technology can be applied to underwater sensors position. In this paper, a CS network is established based on the energy attenuation model of underwater acoustic signals, the position information of the target is taken as the sparse vector. The area where multiple acoustic source target nodes are located is divided into several grids. Several receiving nodes whose positions are known are placed in the area. The measurement matrix is constructed by the channel characteristics between the receiving nodes and the center of the grids, and the measurement matrix is orthogonalized for satisfying the Restricted Isometry Property (RIP). The observation matrix is composed of RSS ratios received by different receiving nodes according to ergodicity principle. The simulation results show that when the position information is not complete, the positions of multiple targets can be obtained simultaneously.