Reactive Motion Control of Humanoid Robot Based on Human Intuitive Intention
碩士 === 國立臺灣大學 === 電機工程學研究所 === 106 === Human robot interaction is a popular topic recently. The implementation of in-teraction not only can use language communication between human and robot but also can use nonverbal intention. Most of research in this region usually adopt mobile robot as the agent...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | en_US |
Published: |
2018
|
Online Access: | http://ndltd.ncl.edu.tw/handle/f2yes8 |
id |
ndltd-TW-106NTU05442041 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-TW-106NTU054420412019-05-30T03:50:44Z http://ndltd.ncl.edu.tw/handle/f2yes8 Reactive Motion Control of Humanoid Robot Based on Human Intuitive Intention 基於人類自然意圖之類人形機器人反應動作控制 Pei-Chun Zheng 鄭培淳 碩士 國立臺灣大學 電機工程學研究所 106 Human robot interaction is a popular topic recently. The implementation of in-teraction not only can use language communication between human and robot but also can use nonverbal intention. Most of research in this region usually adopt mobile robot as the agent. Here we proposed a novel system in anthropomorphic robot. It can realize more friendly interaction with human intention. The process of human robot interac-tion can be divided into two steps. It includes perceiving human behavior from sensors and generating corresponding actions. The Kinect sensor is used for human skeleton tracking which extracts human motions and each foot of subjects is equipped with the IMU sensor to get orientation of each foot more accurately. It can help us get more re-liable information by skeleton extraction. In the same time, the VLAD method men-tioned in this paper can help us to get the mean of human intention. In the experi-mental results, the accuracy of human action classification is 94% which prove the great performance of our method. We also demonstrate the system which is used to effectively generate corresponding actions by mimicking various motion. This paper proposes a novel method for master-slave system of humanoid robot which implements whole body imitation of human motions using humanoid robot de-veloped in our lab in real time. The Kinect sensor is used for human skeleton tracking which extracts human motions and each foot of subjects is equipped with the IMU sensor to get orientation of each foot more accurately. There have some important is-sues are mentioned below. First, in order to ensure stability, we propose sensor fusion in IMU sensor and vision sensor. It can help us get more reliable information by skel-eton extraction. Second, a new dynamic balance strategy which can help the robot im-itation more stable in real time is presented. In the experimental results, we demon-strate the system mimicking various motion effectively. 羅仁權 2018 學位論文 ; thesis 79 en_US |
collection |
NDLTD |
language |
en_US |
format |
Others
|
sources |
NDLTD |
description |
碩士 === 國立臺灣大學 === 電機工程學研究所 === 106 === Human robot interaction is a popular topic recently. The implementation of in-teraction not only can use language communication between human and robot but also can use nonverbal intention. Most of research in this region usually adopt mobile robot as the agent. Here we proposed a novel system in anthropomorphic robot. It can realize more friendly interaction with human intention. The process of human robot interac-tion can be divided into two steps. It includes perceiving human behavior from sensors and generating corresponding actions. The Kinect sensor is used for human skeleton tracking which extracts human motions and each foot of subjects is equipped with the IMU sensor to get orientation of each foot more accurately. It can help us get more re-liable information by skeleton extraction. In the same time, the VLAD method men-tioned in this paper can help us to get the mean of human intention. In the experi-mental results, the accuracy of human action classification is 94% which prove the great performance of our method. We also demonstrate the system which is used to effectively generate corresponding actions by mimicking various motion.
This paper proposes a novel method for master-slave system of humanoid robot which implements whole body imitation of human motions using humanoid robot de-veloped in our lab in real time. The Kinect sensor is used for human skeleton tracking which extracts human motions and each foot of subjects is equipped with the IMU sensor to get orientation of each foot more accurately. There have some important is-sues are mentioned below. First, in order to ensure stability, we propose sensor fusion in IMU sensor and vision sensor. It can help us get more reliable information by skel-eton extraction. Second, a new dynamic balance strategy which can help the robot im-itation more stable in real time is presented. In the experimental results, we demon-strate the system mimicking various motion effectively.
|
author2 |
羅仁權 |
author_facet |
羅仁權 Pei-Chun Zheng 鄭培淳 |
author |
Pei-Chun Zheng 鄭培淳 |
spellingShingle |
Pei-Chun Zheng 鄭培淳 Reactive Motion Control of Humanoid Robot Based on Human Intuitive Intention |
author_sort |
Pei-Chun Zheng |
title |
Reactive Motion Control of Humanoid Robot Based on Human Intuitive Intention |
title_short |
Reactive Motion Control of Humanoid Robot Based on Human Intuitive Intention |
title_full |
Reactive Motion Control of Humanoid Robot Based on Human Intuitive Intention |
title_fullStr |
Reactive Motion Control of Humanoid Robot Based on Human Intuitive Intention |
title_full_unstemmed |
Reactive Motion Control of Humanoid Robot Based on Human Intuitive Intention |
title_sort |
reactive motion control of humanoid robot based on human intuitive intention |
publishDate |
2018 |
url |
http://ndltd.ncl.edu.tw/handle/f2yes8 |
work_keys_str_mv |
AT peichunzheng reactivemotioncontrolofhumanoidrobotbasedonhumanintuitiveintention AT zhèngpéichún reactivemotioncontrolofhumanoidrobotbasedonhumanintuitiveintention AT peichunzheng jīyúrénlèizìrányìtúzhīlèirénxíngjīqìrénfǎnyīngdòngzuòkòngzhì AT zhèngpéichún jīyúrénlèizìrányìtúzhīlèirénxíngjīqìrénfǎnyīngdòngzuòkòngzhì |
_version_ |
1719195360041107456 |