Visual Servoing Meets LQR: Streamlining Automated Cliche Mounting with Computer Vision and Precise Three-Axis Control

The current state of the art in plate mounting techniques relies on the iterative refinement of successive approximations which present a significant challenge in time required to converge on an optimal solution along with its rigidity that does not allow an easy managments of precision-speed tradeoff. The problem is handled with the introduction of an linear quadratic regualtor which is a type of controller that relies on parameters that can be choosen with a good amount of freedom, in order to change the systtem behaviours. Moreover, in the current landscape, visual recognition systems in plate mounters rely on robust pattern recognition technologies to achieve accurate plate alignment. The system relies on low-level functions, which can encounter challenges with changes in luminosity and may not consistently handle plate imperfections with robustness. The increasing computational power of nowadays computers allows to use more complex machine vision programs, so in this thesis there are Implement high-level functions for image recognition and machine vision to improve the stability of the visual component across various operating conditions. Using the LQR contral strategy and the optimized vision part the program is able to reach the target postion with a negligible error under high final error state weights while increasing the velocity by many times mantaining the final error very small with low wights.