Developing a machine learning-based drag-and-play system for the automatic synthesis of optimal and correct-by-construction schedules

Abstract

Due to the rise of Industry 4.0, flexible manufacturing systems and automation solutions with machine learning solvers have been widely adopted by manufacturers to provide flexibility in the assembly line. With the operation of Automated Guided Vehicles (AGV) based on the Discrete Event System (DES) framework in a flexible manufacturing system, route optimization techniques have been used to improve its scheduling performance. However, due to the complexity Vehicle Routing Problem (VRP), several constraints under given conditions have to be considered to reach an optimal solution. By considering the various constraints in VRP, an analysis of the AGV system can be done to improve efficiency. In this paper, we will discuss and experiment with the application of control theories and machine learning techniques to optimize logistic transportation for an AGV system using Google Optimization Tools (OR-Tools). Visualization of AGV routing in the assembly line will be conducted using a 3D simulation program, Visual Components. With the visualization, OR-Tools with simple machine learning techniques will account for the constraints to strategize an optimal route for AGV. Keywords: Machine Learning, Automated Guided Vehicle (AGV), Discrete Event System (DES), Vehicle Routing Problem (VRP), Google Optimization Tools (OR-Tools), Visual Components