Robotic inspection can offer competitive advantages to industrial inspection. First it can allow for keeping humans out of harms way. Robots can undertake most dirty, dull and dangerous jobs. Simultaneously, robotic inspection can lead to more efficient inspection results thus reducing the downtime of industry facilities. This in turn means reduced costs and increased profits. But these goals depend upon two core abilities for robotic systems, namely a) they need to be able to navigate and traverse any kind of complex industrial facilities, and b) they need to be able to autonomously map and reason about their environment in a manner that is analogous to that of humans by detecting and understanding the role of different objects and entities in the environment. To facilitate these ambitions, this project seeks to develop a new generation of walking and flying robotic systems, including their marsupial combination, that can handle diverse industrial environments and conduct autonomous inspection operations by reasoning for every object of interest in a manner tailored to the diverse inspection needs and goals. Accordingly, not only will humans have better means to conduct inspection operations safely and remotely but will also reduce the need to conduct teleoperation or go over excessive amounts of data. Instead, the robotic team will handle the mission autonomously and point out the areas of the industrial environment that possibly require the special attention of human operators. To achieve these exciting goals, we have established a partnership of academia and industry with a clear focus on fundamental research combined with challenging experiments in real-life conditions.
Inspection of industrial facilities is essential in order to maintain them in proper condition, maximize their lifespan, improve production quality, boost profits and reduce their negative impact on the environment. These tasks are still primarily undertaken by humans, occasionally assisted by remotely-controlled robotic systems, as an inspection mission involves complex navigation, scene reasoning, and interaction tasks, alongside correlation with prior data and expert knowledge. This, however, entails that such missions are often very time-consuming, thus leading to significant downtime for the industry and reduced production, while human inspectors face hazardous working conditions. Robotics have for years tried to assist and partially have done so: it is now not uncommon to see a quadrotor - for example - being piloted by an operator on-site and used to capture images. Yet, this merely means that human inspectors have better tools for their job, while the bulk of the activities remains extremely manual.
A viable and appealing alternative is to automate this process in an end-to-end manner, a task only possible if robots act fully autonomously in undertaking all the complexities of inspection missions. This includes traversing and accessing extremely challenging environments, reasoning about the objects in their environment, representing the environment in a manner that allows comparison to previous inspections, and performing large-scale missions in a short amount of time thus reducing downtime. Responding to this task, this project aims to research and develop the breakthroughs necessary towards collaborative robotic teams of walking, roving and flying systems-of-systems that go beyond mapping the environment and facilitate semantic reasoning and characterization, and semantically-driven inspection path planning, with full autonomy. To exploit the technologies within a sector of priority, we focus on the energy and oil & gas industries in Norway and the world.