Image-based Feedback Control for Drift Compensation in an Electron Microscope

Electron microscopes are high-precision imaging systems that are used to image objects (specimens) up to atomic resolutions. Even in highly-controlled industrial sites with minimal external disturbances, internal disturbances cause the objects to move in the field of view of the imaging system. These internal disturbances are jointly considered as drift, which can prohibit fast and accurate imaging. The objective of this paper is to design a model-based control system that leads to the stabilization of the region of interest of the object in the center of the field of view despite the presence of drift, which can be achieved by controlling the movement of the object itself using an actuated positioning system (stage). An interesting feature of the problem is that the displacement of the object can only be measured through the imaging system itself, which requires a non-trivial image-processing step affected by significant delay and inaccuracy. In the image-processing step, a tracking algorithm provides drift estimates. The image-processing characteristics inherent to the tracking algorithm such as image-acquisition time, tracking delay, and measurement accuracy, are explicitly taken into account in a novel LQG-type control design for this application. The decision for a LQG-type control design is based on a stochastic model proposed for the drift motion of the specimen to include the uncertainty in the drift. Moreover, we show that new switched control designs can improve our original (non-switched) LQG design even further. To validate this approach, a simulator has been developed in close collaboration with an industrial OEM of electron microscopes to simulate a relevant industrial case study including tracking software and system parameters used in the industry. The simulator automatically generates and analyzes representative images to simulate closed-loop imaging and processing for feedback control.