Glossary · AI & ML
Imitation learning
In brief
Imitation learning is training a robot policy by showing it human demonstrations rather than by reward shaping. The robot watches a teleoperator (or a human directly), then learns to reproduce the trajectory or behavior conditioned on observations.
Imitation learning bypasses the reward-engineering problem at the heart of reinforcement learning. Instead of carefully crafting a reward function and waiting for the policy to discover good behavior, you collect demonstrations of the desired behavior and train the policy to mimic them. Behavioral cloning is the simplest form; methods like DAgger, GAIL, and diffusion-policy variants address the distributional shift that makes naive cloning brittle.
For humanoids, imitation learning is the dominant data-collection path right now: teleoperators wear a haptic suit or use a VR rig, control the robot through the task, and the trajectory is logged. Once tens of thousands of trajectories are collected per task, a policy is fit and refined.
Related terms
See it in the wild
Browse robots and brands using these techniques
Glossary entries are upstream. The catalog is where the implementations live.