Controller design and optimization for large-delays image processing in visual closed-loop systems

摘要

This paper details the design and optimization of a control system to overcome the problems caused by large-delays in image processing during the automatic focusing of an optical microscope. Two controller designs are simulated and compared. The first design tries to predict the information in between captured images to provide real time feedback. The second design uses a feedforward approach to minimize error. The designs are also compared to a baseline PID controller and both are found to reduce tracking error by as much as 80% in simulation. The predictive controller offers slightly better performance but the feedforward system is much simpler to implement and it is chosen as the best controller for this application. The control system is tested on a scientific instrument that simulates human alveoli. A feedforward model is implemented in the experimental system. Experiments showed satisfactory results and automatic focusing was achieved on a bubble that was cycled by continuously being inflated and deflated to simulate breathing.