Autonomous mobile robots: configuration of an automated inspection system

Robotic technologies are becoming popular and almost essential nowadays in many industrial sectors and for the most disparate purposes. They provide huge benefits to companies and workers: carry out monotonous, unpredictable, and hazardous jobs in harsh environments, resulting in an increased safety and availability of personnel that can be differently deployed. Plant inspection and maintenance are generally human-intensive. Tedious and repetitive tasks may entail decreasing attention spans of employees, leading to errors that may have serious consequences. In that context, robots accomplish thoroughness, safeness, cost and time efficiencies. This thesis work is part of the project carried out by Sprint Reply from the group Reply S.P.A., a company hard-experienced with Robotic Process Automation, OCR tools, Natural Language Processing as well as physical social robots. The project aims to develop a robotic-aided solution for a world leader company in the Oil & Gas industry, to detect gas or liquid leakages, monitor the conditions of working machines, check the availability of HSE equipment and create a digital representation of the sites using 3D LIDAR scans. Spot from Boston Dynamics is the robot used to perform the required tasks. It is an agile mobile robot that allows to automate routine inspection tasks of any area of interest, to monitor different scenarios, detecting and recognizing specific objects and capturing data safely, accurately, and frequently. Using the Software Development Kit, it is possible to create custom controls, program missions, and integrate sensor inputs into data analysis tools. In particular, the thesis activity focuses on visual data processing that consists in building image processing algorithms. During scheduled patrols in the facility plant, Spot recognizes target elements through a widely used neural network for object detection and takes photos that are processed using computer vision. The algorithms recognize peculiar characteristics in the acquired images and verify if they match with a set of given parameters. They collect information and produce as output numerical values or alert messages that are communicated to the operators, visible through a management console.