计算机科学与探索 ›› 2020, Vol. 14 ›› Issue (8): 1275-1287.DOI: 10.3778/j.issn.1673-9418.1912025

• 综述·探索 • 上一篇    下一篇

机器人演示学习编程技术研究综述

殷聪聪,张秋菊   

  1. 1. 江南大学 机械工程学院,江苏 无锡 214122
    2. 江苏省食品先进制造装备技术重点实验室,江苏 无锡 214122
  • 出版日期:2020-08-01 发布日期:2020-08-07

Review of Research on Robot Programming by Learning from Demonstration

YIN Congcong, ZHANG Qiuju   

  1. 1. School of Mechanical Engineering, Jiangnan University, Wuxi, Jiangsu 214122, China
    2. Jiangsu Key Laboratory of Advanced Food Manufacturing Equipment and Technology, Wuxi, Jiangsu 214122, China
  • Online:2020-08-01 Published:2020-08-07

摘要:

传统的工业机器人编程方式对从业人员的编程水平提出了较高的要求,并且编程周期较长,难以满足多品种、小批量、短周期的生产要求。因此,基于演示学习(LfD)的机器人编程技术的研究逐渐兴起。首先,介绍了机器人演示编程技术的发展背景,给出了演示编程技术的定义;然后,按照机器人所获取信息的逻辑层次将演示学习技术分为基于运动和基于任务两类;接着,围绕近年来演示学习编程技术的研究进展与成果,分析了不同种类的演示学习编程技术的原理与特点,综述了演示学习编程技术的研究现状及目前存在的问题;最后,探讨了演示学习编程技术面临的挑战,指明了演示学习智能化、通用化的发展趋势。分析表明,现阶段演示学习编程仍然面临稳定性、通用性、易用性等挑战,机器人演示学习编程对大幅降低工业机器人的编程难度,提高工业机器人的编程效率,以及提升服务机器人的人机交互体验都具有重要意义。

关键词: 演示学习, 机器人编程, 行为识别, 动作分割

Abstract:

The traditional industrial robot programming method puts forward higher requirements for the programming level of employees, and the programming cycle is long, which is difficult to meet the production requirements of multiple varieties, small batch and short cycle. Therefore, the research on robot programming technology based on learning from demonstration (LfD) gradually emerges. Firstly, the development background of robot programming by demonstration technology is introduced, and the definition of robot programming by demonstration is given. Then, the learning from demonstration technology is divided into two categories according to the logical level of the information obtained by the robot: motion-based and task-based. Then, focusing on the research progress and achievements of robot programming by demonstration in recent years, the principles and characteristics of different kinds of programming by demonstration are analyzed, and the research status and existing problems of programming by demonstration are summarized. Finally, the challenges of programming by demonstration are discussed, and the intelligent, universal trend of learning from demonstration is pointed out. The analyses show that the present program-ming by demonstration still faces the challenges of stability, universality and ease of use, and it is of great signi-ficance to greatly reduce the programming difficulty of industrial robots, improve the programming efficiency of industrial robots, and improve the human-computer interaction experience of service robots.

Key words: learning from demonstration, robot programming, action recognition, motion segmentation