QQ客服 官方微博 反馈留言

	Learning Basic Unit Movements with Gate-model Auto-encoder for Humanoid Arm Motion Control
	Hu, Fan ; Liu, Wentao ; Wu, Xihong ; Luo, Dingsheng
	2016
关键词	Humanoid Arm motion control Basic unit movements Deep neural network
英文摘要	Manipulation is one of the most important skills for humanoid robots and is a hot topic in robotics. To acquire the manipulation skill, robots need to know how to control its arm to drive the hand to the desired position, which is called arm motion control. In our research, we try to equip our robot with the ability to accomplish complicated manipulation tasks by combining basic unit movements (BUMs), which makes it easier and more robust. In exploring the method to learn the BUMs, we have proposed a Deep Neural Network (DNN) based method with which the robot is able to execute BUMs in the whole workspace of his right arm with a high accuracy. However, the chosen BUMs are unpractical for that they are not mutually independent, and they have no attribute of amplitude. So in this paper, we firstly propose a stricter and more practical definition of BUMs to ensure the independence of BUMs. After this, in order to learn BUMs under the improved definition as well as to learn the amplitude at the same time, we changed our original network (classical auto-encoder, CAE) into a new model (gate-model auto-encoder, GMAE) to adapt to the new BUMs. We conducted experiments to compare the performance of CAE and GMAE, the results on both simulation and the PKU-HR5II robot have proved the effectiveness of the proposed methods.; National Basic Research Program (973 Program) of China [2013CB329304]; National Natural Science Foundation of China [11590773, 61421062]; Key Program of National Social Science Foundation of China [12 ZD119]; CPCI-S(ISTP); 246-251
语种	英语
出处	IEEE International Conference on Information and Automation (ICIA)
内容类型	其他
源URL	[http://ir.pku.edu.cn/handle/20.500.11897/470105]
专题	信息科学技术学院
推荐引用方式 GB/T 7714	Hu, Fan,Liu, Wentao,Wu, Xihong,et al. Learning Basic Unit Movements with Gate-model Auto-encoder for Humanoid Arm Motion Control. 2016-01-01.

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。