A new research initiative announced today by the University of California hopes to bring together cloud robotics, robot learning, human-centric automation and bio-inspired robotics, with the ambition to reduce drudgery and improve the human experience in sectors where robots are destined to prove most useful.
For decades, robots have been contributing to our standard of living by working in manufacturing and logistics operations, contributing to productivity by being reliably repetitive. However, getting robots to work in less structured conditions, where they have to deal with variability and understand how to adapt their actions to the environment, has been very difficult.
Created at the Center for Information Technology Research in the Interest of Society (CITRIS), the new multidisciplinary, multi-campus People and Robots Initiative plans to change that. “People and Robots is not one of those ideas that is 20 years away indefinitely. New, human-safe robots are already available for manufacturing, robot-assisted surgery has been used for millions of procedures,” says roboticist Matei Ciocarlie at Columbia University in New York. “This initiative should accelerate progress in all these areas and hopefully open up new ones where robots aren't used yet.”
The scope of the research will range from projects such as human-guided robot learning, disaster preparedness and disaster response to precision watering for agriculture and home decluttering.
Researchers will also study how people work together with machines to solve difficult problems – the so-called ‘multiplicity’ concept. “Multiplicity combines emerging results in collective intelligence and cloud computing, building on research in ensemble learning, big data, open-source software, and industry initiatives,” says roboticist Ken Goldberg of University of California, Berkeley (UCB), who is part of CITRIS. “This project brings together some of the world's top roboticists to advance the cutting edge of research that specifically benefits humans and involves humans in the loop.”
The latest cutting-edge technologies will come into play with People and Robots, such as collective intelligence, cloud computing, big data, open-source software and the Internet of Things, as well as research in psychology, law, ethics, art and the humanities, to help address human issues such as inclusion, privacy, and alienation.
First created in 2001, CITRIS has a long history of studying IT and its impact on society. “With the new initiative they are including robotics as a major component and also the impact of cloud robotics on manufacturing, healthcare and services,” notes Henrik Christensen of Georgia Tech in Atlanta, Georgia, who is not part of CITRIS.
At Georgia Tech, there is a similar initiative at the Institute on Robotics and Intelligent Systems with the emphasis there on people-centred robotics, or ‘co-robots,’ as they are known at the faculty.
The fastest-growing area of robotics today is collaborative robots, where people and robots work side-by-side on manufacturing lines or at e-commerce shops such as Amazon. There are also robots beginning to be used as assistants in homes or for mobility support to people with disabilities. “We are seeing a confluence of materials, design, computing, control and interaction that opens new worlds of interaction and assistance,” says Christensen. “Over the next five to ten years we will see a serious impact on restoring manufacturing, enabling improved quality of life for an ageing population and as part of the new services such as delivery.”
CITRIS hopes to achieve this by addressing a number of fundamental problems with robotics. “What's difficult for robots today is to generalise to new situations. And that's exactly where our work comes in,” says Pieter Abbeel, a colleague of Goldberg at UCB. “If we can equip robots with general skills, they can readily be deployed for a wide range of tasks, such as cleaning homes, organising laundry, cooking and so on in our everyday environments, which have high variability and will never match what the robot got to see in a lab.”
To do this, rather than equipping robots directly with every possible skill, the researchers want them to have the ability to learn to pick up new skills on their own, either from watching people (or videos of people) performing certain tasks or through trial and error. The initiative will use computer vision and speech recognition and aims to “examine how we can advance deep learning even further to enable similar significant leaps forwarding robot learning capabilities,” says Abbeel.
Initial trials have already shown that it is possible for robots to learn to perform challenging manipulation tasks, such as placing blocks in to matching openings or screwing caps on to bottles. “What's so intriguing about these results is not that we now have a robot that can do these tasks. What’s so intriguing is that we now have a learning approach that can equip a robot with new manipulation skills that require hand-eye coordination in a very general way,” says Abbeel.
It’s also not about purely abstract scientific advances. “Engineers take pride in building solutions to real problems and devices that are useful to people. I believe this initiative relies on engineering at its core and will also advance engineering in order to build physical robots better at interacting with people,” says Ciocarlie.