We address a self-organizing network problem for autonomous robotic sensor swarms towards various ad hoc sensor network applications. For this purpose, a decentralized solution approach is proposed to enable individual robots to build and to maintain their network through locally communicative interactions with adjacent robots. Specifically, the communicative interaction allows robots to select specific neighbors with higher connectivity, and to adapt to network topological changes by robot movements. The proposed approach is verified to be effective through extensive simulations for the secure self-organization of a robotic sensor network. Considering the realistic conditions of communication delays and re-configuration for robot disappearances due to robot failures, simulations are also performed.