A semi-automatic multi agent intelligent system for production processes

With the expansion of the industry 4.0, the factories, with the aim of improving production processes, are more and more focusing on connectivity and interaction among machines and people. Multi agent systems are an example of communication between software agents and they are widely used in different domains. The objective of the thesis is to design, develop and test a multi-agent intelligent system in the context of industry 4.0. Nowadays there exists a variety of multi-agent system framework, some of them are flexible and powerful, but usually they lack of the intelligent part, the proposed platform already integrates a machine learning algorithm that can be used by the agents to take actions based on the data they are perceiving. Moreover, due to the low risk policy affecting the industry scenario, starting from a complete automation of the process is too venturesome and dangerous. So, this system provides a partial automation exploiting a human-in-the-loop model, in this way the operator has a complete control on what is happening on the plant. Each agent’s action must be validated by the supervisor before it takes effect and it is also a way to improve agent’s decision. The system was finally tested in a partially simulated environment with an unsupervised dataset, in order to show the ability of the agents to communicate among themselves and the ability to interact with the human supervisor.