“We have won the biggest worldwide competition in Robotics and AI”

The research group in Robotics and AI at Luleå University of Technology is world class in robotics, not the least proven in the Subterranean Challenge. From left: Christoforos Kanellakis, Sina Sharif Mansouri, Anton Koval, Jakub Haluska, Andreas Papadimitriou, George Nikolakopoulos, and Björn Lindqvist. Photo: Linda Alfredsson

The Robotics and AI research group continues its successful journey in NASA/JPL’s team CoSTAR in the Subterranean Challenge, the biggest robotics competition in the world. The team won the second circuit of the competition and is now preparing for the third and last overall combined challenges.

– We have won the biggest worldwide competition in Robotics and AI and it is an honor for us to collaborate with NASA/JPL, Caltech, MIT and KAIST. It is a constant source of inspiration and it is amazing that our research group is participating in a research movement that will change the application of AI through robotic technologies in all our aspects of our life in the next 5–7 years, says George Nikolakopoulos, Professor of Robotics and Artificial Intelligence.

The second stage of the competition was focusing on urban underground environments. It took place in an unfinished nuclear power plant in the state of Washington, USA. The task was similar to the first circuit – the robots should navigate and explore autonomously, the overall aim being to detect and locate artifacts, for example specific objects and humans, and report the artifact’s positions. This time the robotic technologies deployed were mainly based on the so-called Spot, a quadruped robot manufactured by Boston Dynamics.

– The mission was a typical search and rescue mission. We put robots in unknown and unstructured environments where advanced AI is necessary for enabling a full autonomous mission, says George Nikolakopoulos.

– This competition is really pushing the limits of AI since the robots must operate in very difficult environments and with a very high level of robustness and adaptation to the corresponding needs.

Difficult algorithms

Besides NASA/JPL, team CoSTAR consists of the California Institute of Technology, Massachusetts Institute of Technology, and Korea Advanced Institute of Science and Technology. In total, eleven teams are in the competition, they are all top participants from the area of robotics and autonomous systems from all around the world. Since the start of the competition, the Robotics and AI team has contributed with their experiences in Field Robotics and deploying autonomous robots in real life environments. It is a big challenge to achieve high levels of autonomy and robustness of the software’s algorithms; it is difficult to create algorithms that are capable of operating all the time with the same performance.

– Our contributions are specifically in the field of aerial robotics and from an algorithmic perspective. We focus on path planning, reactive navigation for avoiding obstacles, environmental perception, and 3D reconstruction, says Postdoc researcher Christoforos Kanellakis.

The Subterranean Challenge consists of three challenges. The participant teams are supposed to address autonomy, perception and mobility technologies needed to map subsurface networks in unpredictable conditions.

The first challenge was to map, identify, and report artifacts in a mine by utilizing autonomous systems. In that circuit CoSTAR – the team to which Luleå University of Technology belongs  – finished second.

The second challenge was about navigating in an urban underground environment and the team finished first. The third challenge will take place in October, followed by a final event in August next year.