There's much robots can achieve by observing human demonstrations, like the actions necessary to move a box of crackers from a counter to storage. But imitation learning is by no means a perfect science -- demonstrators often complete subgoals that distract systems from overarching tasks. To solve this, researchers at the University of Washington, Stanford University, the University of Illinois Urbana-Champaign, the University of Toronto, and Nvidia propose an "inverse planning" system that taps motions or low-level trajectories to capture the intention of actions. After evaluating their technique by collecting and testing against a corpus of video demonstrations conditioned on a set of kitchen goals, the team reports that their motion reasoning approach improves task success by over 20%. The researchers lay out the full extent of the problem in a preprint paper detailing their work.

It takes skill to fall with grace. Not in a metaphorical sense, but in a very real one: emerging relatively unharmed after a tumble is a skill human have and can refine that robots lack innately, often to comical effect. New research by Georgia Tech wants to protect expensive, wobbly robots of the future, by programming them to fall safely and gracefully. There are plenty of robots designed to avoid falling, like Alphabet's surreal animal-like BigDog series of machines. But bipedal robots, or bipedal-ish robots, designed to operate in spaces made for humans, don't have the luxury of a weird, well-balanced body.

Researchers from Germany are developing an artificial nervous system aimed at teaching robots how to feel pain. As well as allowing robots to quickly respond to potential damage to their systems, it could also protect humans who are increasingly working alongside them. The scientists plan to base the system on "insights from human pain research". To test it, they fitted a robotic arm with a fingertip sensor that could detect pressure and temperature. The researchers, from Leibniz University in Hannover, are developing a system that would allow a robot to "be able to detect and classify unforeseen physical states and disturbances, rate the potential damage they may cause to it and initiate appropriate countermeasures, ie reflexes", they explained.