Phase-based skill learning improves robot latte art pouring accuracy
One-Shot Learning for Double-Shot Latte Art: Adapting Cobot Pouring Skills with Dynamic Movement Primitives
Robotics
Summary
Making latte art needs very precise and careful movements, which can be hard for robots to learn. The authors used a method where a single demonstration from a barista is broken down into separate parts, each learned individually, making it easier for the robot to copy. This approach greatly improved the robot’s accuracy in pouring and tilting the jug and allowed it to adjust to different cup sizes without needing a new demonstration. The method can also be used for other tasks with clear phases that need different movements.
What this means in practice
- •For roboticists in food service: Program cafe robots to learn complex pouring tasks from a single demonstration with phase-based motion encoding.$Commercial implications: Enables development of robotic baristas capable of replicating skilled latte art recipes with minimal training, improving product quality and customization.
- •For industrial robot programmers: Adapt robot actions for tasks with distinct phases by encoding phases separately to improve precision and scalability.
Authors
Zain Khan, Victor Hernandez Moreno, Dinh Tung Le
Abstract
Latte art is a complex task requiring dexterity and precision, making it a challenging and commercially relevant task. A one-shot learning system is employed in which a single barista demonstration is segmented into five functionally distinct phases, each encoded as an independent Cartesian Dynamic Movement Primitive (CDMP) with its own parameters. Against a single-CDMP baseline, per-phase encoding reduces transition endpoint error by 3.2-4.5 times and art-stroke shape error by 2.7 times, reproducing jug tilt to below 1 degree RMS in the pouring phases. The learnt pour scales from a 280mL to a 400mL cup through a single spatial offset and a scaled recipe, with no re-demonstration. The phase-based CDMP approach generalises beyond latte art to any demonstration task whose phases have distinct spatial or temporal requirements whilst demonstrating the ability to adapt reproductions to suit different task requirements.