Robots learn to move expressively while staying balanced and interactive
RoboMMT: Learning Stable Tracking of Modulated Expressive and Interactive Robot Motion
Robotics
Summary
Robots that move expressively can appear more natural and friendly, but this can make them unstable, especially on tricky walking platforms. The authors created a new system called RoboMMT that helps robots combine expressive motions with balance and goal-driven reactions. It learns how to adjust movements in real-time to stay steady while interacting with people or the environment. They showed this working on different robot types, including ones that walk on four legs and two legs, even when pushed or disturbed.
What this means in practice
- •For robotics developers: Build interactive robots that perform expressive body motions while maintaining balance in complex real-world environments.
- •For animation engineers: Create physics-based animated robot characters that can react to user commands while preserving stable and natural movement.
Authors
Till Hielscher, Kai O. Arras
Abstract
Expressive robot motion can improve naturalness, effectiveness, and perceived likability in human-robot interaction. When deployed on dynamically unstable platforms, however, reactive execution of expressive body motion may bring the robot to its stability limits, especially when expressivity is decoupled from physical control. In this paper, we present RoboMMT (Robot Modulated Motion Tracking), a modular control architecture that couples parameterized expressive motion synthesis, physics-based character control, and reactive adaptation to spatial interaction goals. Our approach trains a reference-conditioned tracking policy across a broad domain of expressivity, enabling it to faithfully realize the commanded motions while maintaining dynamic stability. Goal reactivity is achieved by training lightweight, task-specific interaction policies that modulate the reference generator's spatial goal-attraction point, with the stabilizing tracking policy kept in the loop. Evaluations on a quadruped and two humanoid robots show that the learned tracker remains stable and accurate across the entire modulation domain, whereas trackers trained without this coverage lose balance, tracking accuracy, or both. We show that layering closed-loop goal adaptation over this foundation enables interactive facing and pointing behaviors, while maintaining stable execution and comparable reference-tracking fidelity. The robot learns to proactively redistribute its weight and take corrective steps. Zero-shot sim-to-real transfer is demonstrated through real-robot experiments on a quadruped under heavy disturbances.