BROWSE
Detail
Conference Paper
Deep Learning-based Behavior Generation for Social Robots
Cited - time in 
Share 
Share
- Authors
- Woo-Ri Ko, Jaeyeon Lee, Jaehong Kim, Minsu Jang
- Issue Date
- 2018-03
- Citation
- International Conference on Human-Robot Interaction (HRI) 2018 : Workshop, pp.1-3
- Language
- English
- Type
- Conference Paper
- Abstract
- This paper proposes a reactive behavior generation method for social robots using deep neural networks. Low-dimensional features are extracted from the robot's point-of-view images by using YOLO detector. Then, considering the extracted image features and current joint angles, the robot's next joint angles are generated by long short-term memory (LSTM)-based behavior generator. To show the effectiveness of the proposed method, a small humanoid robot, "NAO, " is trained to perform a bell-ringing task under direct teaching.
- KSP Keywords
- Behavior Generation, Deep neural network(DNN), Image feature, Joint angles, Learning-based, Long-short term memory(LSTM), Low-dimensional Features, Point-of-view, deep learning(DL), humanoid robot, social robot