Many different kinds of robot systems have been successfully deployed in complex environments, while research into collaborative control systems between different robots, which can be seen as a hybrid internetware safety-critical system, has become essential. This paper discusses ways to construct robust and secure reliability architecture for collaborative robot control systems in complex environments. First, the indication system for evaluating the real-time reliability of hybrid internetware systems is established. Next, a dynamic collaborative reliability model for components of hybrid internetware systems is proposed. Then, a reliable, adaptive and evolutionary computation method for hybrid internetware systems is proposed, and a timing consistency verification solution for collaborative robot control internetware applications is studied. Finally, a multi-level security model supporting dynamic resource allocation is established.