mariofiorino/master-thesis

Final work of the master's degree course in Artificial Intelligence and Robotics - Sapienza, Rome

Master-Thesis

Dealing with anomalies in robot-enabled industrial quality control laboratory - Mario Fiorino 1871233

In this repository, the final work of the master's degree course in Artificial Intelligence and Robotics - Academic Year 2020-21, Sapienza, Rome

https://sites.google.com/view/mariofiorinothesis/home

This thesis work was born and developed within the European context: AIPlan4EU from the collaboration of the Procter & Gamble Services Company - Belgium, and Sapienza University of Rome.

I had a chance to develop a robotic simulator in Docker container, a framework for AI planning that includes the paradigm "Human-in-the-loop", and designed various AI plans.

This thesis is the starting point of a project that, started on January 2021, will continue for the next three years. This and the works that will be subsequently developed, will be stored and published in the GitHub repository of Prof. Luca Iocchi.

Brief presentation of the chapters

In Chapter 1, motivations and objectives of this thesis work.

In Chapter 2, an introduction to AI planning, the basic ideas of computational learning technique in the field of robotic behaviors, and Learning from Demonstration paradigm are briefly presented. With this in mind, the description of the main papers which have inspired the formation of this thesis work and its future developments are outlined.

In Chapter 3, the software technology used for the building of the robot simulator and its virtual lab environment: Docker, ROS, Gazebo, is described. Their advantages and disadvantages, how they have been used to develop this thesis, and the results obtained are shown. Then are presented an essential description of the main real laboratory tools: Universal Robot 5 and Robotiq 2F-85 gripper, digitized in the simulator. The chapter ends with Petri Net Plans and PlanExecution Interface frameworks, with which the plans were implemented.

In Chapter 4, the nominal plan as a whole; the result of analysis the cases of the fail: their categorization into four areas; and finally the principles, centered around the concept of Human-in-the-loop, underlying the proposed recovery solution are illustrated.

In Chapter 5, the implementation phase of the plans and its results are presented. Starting from the design of simple plans, more complex plans have been formulated,and, at the end, proposed directions and objectives for future works.

In Appendix, a series of technical aspects, used in this thesis, are presented; for the reader who would like to know more.