There is increasing interest in using technological interfaces to explore and expand the capabilities of artificial life research. We are beginning to see systems that exploit the integration of wetware, software and hardware. Here we focus on the overall architecture for real-time model-based control of thermocapillary motion of droplets in a liquid layer with a low power laser beam on a modular robotic platform. We developed a control-oriented class of models using a closed-loop system architecture, that is characterized by state estimation, parameter identification and control. This results in a fast and potentially robust system for the manipulation of far from equilibrium droplet systems in real time.