What is imitation learning in robotics?
Imitation learning trains a robot from demonstrations — teleoperation, motion capture or video — instead of hand-coding every step or relying only on sparse rewards.
Updated:
Definition
Imitation learning (learning from demonstration) teaches a robot by showing examples of a person or teleoperator doing the task, rather than writing every rule or shaping a dense reward.
How it works
Demos arrive via teleoperation, kinesthetic teaching, gloves, or egocentric video. A policy learns to map what the robot sees to what it should do. Scaling that data is now an industry of its own.
Why it matters
It is the main practical path from "robot that waves in a demo" to policies that generalize across rooms — and the reason data startups sit next to model labs in the funding stack.
Frequently asked questions
Is it the same as behavior cloning?
Behavior cloning is the simplest form: map observations directly to actions. Broader imitation learning also includes inverse RL and preference methods.
Why do humanoid labs care?
Homes and factories have too many tasks to script. Large demo datasets feed [VLA models](/{L}/glossary/vla-model).
How does Mecka fit in?
Companies like Mecka scale human motion capture so labs get contact-rich demos for [embodied AI](/{L}/glossary/embodied-ai).
Coverage
Mecka raises $60M Series B to digitize human motion for robot training
Sequoia led Mecka's $60M Series B with NVIDIA, Microsoft's M12, Qualcomm Ventures and Samsung; the startup captures human demonstration data and says it passed $100M run-rate revenue in June 2026.
Newsletter
Robrief Weekly
One email a week with the robotics news that matters. Free, unsubscribe anytime.