polito.it
Politecnico di Torino (logo)

Autonomous Industrial Inspections with Spot by Boston Dynamics: Machine Learning and Computer Vision Techniques for Meter Reading

Pierluigi Compagnone

Autonomous Industrial Inspections with Spot by Boston Dynamics: Machine Learning and Computer Vision Techniques for Meter Reading.

Rel. Tatiana Tommasi, Oscar Pistamiglio. Politecnico di Torino, Corso di laurea magistrale in Data Science And Engineering, 2023

Abstract:

In recent years, robotic technologies are spreading more and more in different sectors and for various purposes, given their generally recognized usefulness. As a matter of fact, the use of robots brings advantages to companies in terms of productivity, efficiency and costs, but also to workers by replacing them in monotonous and dangerous activities. As a result, it increases personnel safety and availability for tasks that strictly require human capabilities. The benefits are even greater considering the adoption of autonomous robots, which are able to perform tasks without requiring human intervention at all. Robotics and automation have promising prospects in the field of industrial inspections, which includes all operations aimed at assessing the state of equipment, infrastructure and industrial plants in general. Traditionally, these inspections are regularly carried out by human operators who walk around the plant and perform various actions, such as observing for signs of wear or taking some measurements, to make sure that everything is running correctly. However, robotic systems alone are not sufficient for the complete automation of some complicated processes performed during inspections, which may require data acquisition and analysis to extrapolate the desired information. In this regard computer vision and artificial intelligence innovations are equally fundamental respect to those in robotic and automation fields. This thesis work fits exactly into the transition process towards the automation of industrial supervision, since it regards a research and development project about autonomous plant inspections. The project, in which I collaborated, was carried out by the company Sprint Reply on behalf of a third-party company leader in the Oil and Gas industry. It includes the study of the state-of-art autonomous visual inspection solutions and the development of concrete use cases within a plant of the customer company. The purpose of the inspection tours defined is the completely autonomous reading of the values reported by some of the analog and digital meters located inside the plant. The robot adopted to implement the autonomous visual inspections is Spot, a quadrupedal robot by Boston Dynamics, equipped with a very powerful camera used to acquire detailed images of the gauges to be read. Then, the numerical value of the measurement is extracted by digitally processing these images, using both deep learning and computer vision techniques. Therefore, the project includes a code development phase, concerning the creation of the algorithms responsible for the digital reading pipeline, which are built by exploiting the images acquired in a data collection carried out at the beginning. The other main activity is the definition of autonomous robotic missions in the plant, which concerns the configuration of Spot and its camera. Finally, extensive tests are performed to evaluate both the reading algorithms and the entire autonomous supervision framework. Furthermore, a detailed analysis of the errors is carried out in order to understand what the main causes of failure are and thus provide insights for future improvements.

Relators: Tatiana Tommasi, Oscar Pistamiglio
Academic year: 2022/23
Publication type: Electronic
Number of Pages: 88
Additional Information: Tesi secretata. Fulltext non presente
Subjects:
Corso di laurea: Corso di laurea magistrale in Data Science And Engineering
Classe di laurea: New organization > Master science > LM-32 - COMPUTER SYSTEMS ENGINEERING
Aziende collaboratrici: SPRINT REPLY S.R.L. CON UNICO SOCIO
URI: http://webthesis.biblio.polito.it/id/eprint/26821
Modify record (reserved for operators) Modify record (reserved for operators)