Researchers at the University of Southern California School of Engineering have created an AI-powered robotic system that can predict how a person will assemble furniture — and offer to help. Such a system could come in handy for people with disabilities – for example, for eating or cooking.
For robots to provide real help, they need to learn what humans know almost instinctively–to predict human needs. To do this, engineers have created a system that can predict what a human will need to assemble a cabinet the next moment – and give it a shelf, fasteners, or tools.
Heramb Nemlekar, the lead author of the work, explained that the goal of the project was to teach the robot to be a handyman – and to help him with secondary functions while a human does the main work.
A robot helps a human assemble an IKEA closet. It sees how the human takes the parts and connects them. It surmises what the human will need next based on previous observations.
In 2018. made famous researchers from Singapore, whose robot self-assembled an IKEA chair. In a new study, the engineers focus on how a robot and a human work together. According to the authors of the study, combining the strengths of a human and a robot can produce the best results.
Their task is to make the robot predict what the human needs based on previous actions. The main problem is that humans can solve the same tasks by performing actions in different order.
But it turned out that by observing people, robots were able to detect patterns in their behavior, since the latter’s behavior still combined around a small number of basic preferences. The AI helped classify people into preference groups based on their actions by processing videos of 20 different people assembling the same closet.
In the current version, the robotic arm’s actions have been programmed manually, but future versions can also learn to work by watching its human partner. Engineers are now working on a new test in which a robot and a human build a model airplane together, a job that requires attention to the smallest detail.