This research focuses on the path planning problem that emerges during the inspection of a Printed Circuit Board (PCB) at semiconductor fabs. The inspection process consists of two elements: the camera and the image processor. The camera visits and captures the predetermined locations on a PCB. The inspection is performed on this image by the backend image processor. The objective of the research is to minimize the makepsan of the inspection of a PCB. A reinforcement learning approach is proposed to solve the problem. Firstly, the 2-Opt heuristic method is adopted. A policy network and a state value network are trained on the paths provided by the 2-Opt in a supervised manner. The policy network provided comparable solutions to the target policy which is the 2-Opt heuristic.
* The author of this thesis is a Global Korea Scholarship scholar sponsored by the Korean Government