Augmented Reality Interface for Constrained Learning from Demonstration
Augmented Reality Interface for Constrained Learning from Demonstration
复制标题
用于从演示中进行约束学习的增强现实界面
DOI：
--
复制
发表时间：
2019
期刊：
影响因子：
--
通讯作者：
Bradley Hayes
中科院分区：
文献类型：
--
作者：
Matthew B. Luebbers;Minjae Connor Brooks;John Kim;Daniel Szafir;Bradley Hayes
- 本文提出了一种新的增强现实（AR）界面的可视化和直接控制的机器人技能学习从示范（LfD）。该系统被设计用于一般机器人手臂操作任务的概念约束学习（CC-LfD）算法，其中LfD系统中的轨迹在学习过程中受到各种约束，以确保正确学习所需的技能。该系统提供了一个交互式的可视化观察和学习机器人轨迹，以及主动谓词约束应用于CC-LfD。在本文中，我们描述了目前的工作系统，帮助用户提供改进的轨迹演示使用AR可视化的约束，以及计划的人类受试者的实验，以评估该系统的有用性。此外，我们还讨论了这项工作的未来扩展，包括使用AR界面来修改现有的轨迹演示。
—This paper presents a novel augmented reality (AR) interface for the visualization and directed control of robot skill Learning from Demonstration (LfD). This system is designed for use with the Concept-Constrained Learning from Demonstration (CC-LfD) algorithm for general robotic arm manipulation tasks, in which trajectories in an LfD system are subjected to various constraints during the learning process to ensure that the desired skill is learned properly. This system provides an interactive visualization of observed and learned robot trajectories, as well as the active predicate constraints applied for CC-LfD. In this paper, we describe current work on a system for helping users give improved trajectory demonstrations using AR visualizations of constraints, as well as a planned human subjects experiment to evaluate the usefulness of this system. Additionally, we discuss future extensions of this work involving using an AR interface to modify existing trajectory demonstrations.