Applying Artificial Intelligence Techniques to the scheduling problem

What is it about?

The job shop scheduling problem in industrial engineering is addressed using Deep Reinforcement Learning (DRL) approaches. Discrete Event Simulation (DES) is crucial for training DRL techniques. However, many DRL techniques have been implemented without a simulation environment. This paper highlights their limitations and presents a numerical experiment demonstrating their ineffectiveness. An agent-based discrete event simulator is also presented.

Featured Image

Why is it important?

The paper presents the Job Shop Simulator (JSS), a discrete event simulation model for a dynamic flexible job shop. JSS is suitable for training and testing artificial intelligence techniques like Deep Learning (DRL), as it allows for the evaluation of state variables without approximations. The simulator's architecture allows for easy implementation of new features, reward functions, and dynamic events. It is a parametric object-oriented simulator developed in Java, making it easier to simulate processes without manually building a different model for each instance. JSS's event-oriented approach makes it integrable with AI agents, making training and testing phases more flexible and less time-consuming.