Learning robot assembly from one human demonstration
State-of-the-Art in Learning-by-Demonstration with Passive Observation for Industrial Assembly Automation
Robotics
Summary
Programming robots for complex factory tasks is hard and slow. This paper reviews ways to teach robots by simply watching a person do a task once. The authors highlight how new methods focus on recognizing the objects involved to help robots quickly learn variations of products. This approach allows more flexible and faster robot setup in factories that make many different items in small amounts.
What this means in practice
- •For manufacturing engineers: Use one-shot demonstration systems to speed up robot programming for new product assembly in flexible factories.
- •For robotics system integrators: Integrate perception architectures focusing on object recognition to enable robots to generalize learned tasks to new variants with minimal retraining.
A survey. It maps existing work.
Authors
David Koetter, Oliver Petrovic, Christian Brecher
Abstract
Learning-by-Demonstration (LbD) enables intuitive robot programming by capturing expert skills, which is crucial for agility in high-mix, low- volume manufacturing. This systematic literature review analyzes passive LbD for industrial assembly processes, focusing on the perception architecture and the generalization of the perceived demonstration. We specifically investigate one-shot approaches where only a single demonstration is required. The review evaluates how systems adapt to new assemblies using this limited data. We identify a shift towards object-centric perception, allowing learned primitives to be transferred to new product variants with minimal training.