Description
Teaching robots precise insertion or assembly with pure RL takes thousands of real-world trials. HIL-SERL provides a framework for robotic manipulation RL combining demonstrations and human corrections.
Its libraries, env wrappers and examples train policies to near-perfect success with little human intervention.
Human-in-loop:Corrections.
Demos:Fast start.
Success:Near-perfect.
Examples:Env wrappers.
Its libraries, env wrappers and examples train policies to near-perfect success with little human intervention.
Features
Human-in-loop:Corrections.
Demos:Fast start.
Success:Near-perfect.
Examples:Env wrappers.
