Human-centered robotic systems open a large field of new applications, both in industry and service contexts. For their interaction with human beings, up to physical collaboration, they rely heavily on computer vision, and more recently on human motion tracking algorithms, which are examples of intelligent components. The complexity resulting from the variety of human behaviors and the combination of intelligent components with robotic collaborative tasks raises the problem of the performance evaluation of the overall system. To support experiment design for performance evaluation of intelligent collaborative robotic systems, we propose an approach combining real-world human motion recordings with numerical simulations of the dynamics of the robotic system with its controller. In this article, we illustrate this approach on the example of the handover task.