Robots Learn to Move Like Us: Disney’s New Trick

Disney researchers have created a new system that helps robots imitate human motion more accurately, even when the robot’s body is very different from a person. The project, called “ReActor, ” uses a two‑layer learning approach that tweaks the original human movement and trains the robot to follow it at the same time. This method solves common problems such as feet sliding on the floor, parts of the robot colliding with each other, or movements that are physically impossible for the machine. Instead of manually adjusting every joint, ReActor only needs a few key links between the human and robot bodies.It then runs physics simulations while applying reinforcement learning to teach the robot how to track the modified motion. The team tested their system in computer models and on real robots, including two humanoid machines and a four‑legged robot. In all cases, ReActor produced smoother, more realistic movements than older techniques. If the research moves forward, it could make robotic characters in movies or theme parks move more naturally, reducing the time spent fine‑tuning each robot. The work is still early, but it shows Disney’s continued interest in blending robotics with storytelling.