This paper proposes an approach to developing a cyber-physical model-driven system that performs robotic distant assembly operations in decentralised manufacturing environment. The system uses an industrial robot to assemble components unknown in advance. The system captures and analyses the components’ silhouettes, and constructs the corresponding 3D models. By the help of the models, the system is able to guide a distant human operator to assemble the real components in the actual robot cell. The results show that the developed system can construct the 3D models and assemble them within a suitable total time, imitating the human behaviour in a similar situation.