Expressive robot motion has been shown to make interactions more natural and likable, but it poses a problem for robots that must stay balanced. The new work, described in an arXiv preprint, introduces RoboMMT, a method for learning stable tracking of modulated expressive and interactive robot motion.
The abstract notes that when expressive body motion is executed reactively on dynamically unstable platforms, it may bring stability risks. RoboMMT appears to address this by learning how to track such motion while maintaining stability, though the specific technical details are not included in the available abstract.
The significance lies in enabling more human-like robot behavior without sacrificing safety or control, particularly for robots on wheels, legs, or other unstable bases. The paper is a step toward making expressive robots practical in real-world settings.