Core Concepts
Augmented Reality framework enables non-roboticists to collect low-dimensional demonstrations for robot imitation learning.
Abstract
Abstract:
Introduces an AR-assisted framework for scalable robot imitation learning.
Empowers non-roboticist users to produce demonstrations using devices like HoloLens 2.
Introduction:
Discusses the limitations of current methods in collecting robot demonstrations.
Proposes a novel AR-based solution to address challenges faced by non-expert users.
Methodology:
Describes the process of collecting demonstrations using AR technology.
Details the key poses detection method to enhance demonstration smoothness.
Experiments:
Conducts experiments on three fundamental robotic tasks: Reach, Push, and Pick-and-Place.
Visualizes collected demonstrations and demonstrates successful real robot replay.
Conclusion:
Presents an AR-assisted framework for scalable demonstration gathering for non-experts.
Acknowledgements:
Acknowledges sponsorship by the National Science Foundation (NSF).
References:
Lists relevant references supporting the research presented.