In collective intelligent robotic environment, robots typically behaves in communication with each other to achieve the global intelligent goal. In general, a collective intelligent robotic system works based on a backbone-less ad-hoc network. For controlling robots in that environment, several issues are raised in the system: action control, environment recognition, robot networking, and system supervisor control. The collective intelligent robot system is quite different from general multi-robot systems in the viewpoint of networking, since it should show intelligence collectively with different types of networked robots. That is, the existing network systems do not work well in the collective intelligent robotic environment. We propose an ad-hoc network system with two essential components, routing manager and topology manager, suitable for the collective intelligent robotic environment.