Smart manufacturing control system based on deep reinforcement learning

Abstract

The base of smart manufacturing contains two main components: cyber-physical systems and the Internet-of-Things (IoT). But the progress in machine learning, Big Data and deep learning technologies have a great impact on all areas of the modern economy, including information technology, industry. The capabilities of deep learning models associated with the ability of intelligent systems to make decisions significantly transform the key paradigms of intelligent production systems. The gap between the principles of human decision-making and intelligent systems is constantly closing. New approaches to decision-making based on deep models allow bringing the benefits of human control into production processes, and, at the same time, getting rid of errors associated with the human factor. The paper proposes an architecture for building multipurpose manufacturing systems based on deep machine learning models and reinforcement learning technologies.