期刊名称：International Journal of Computer Science and Network Security
印刷版ISSN：1738-7906
出版年度：2019
卷号：19
期号：6
页码：91-97
出版社：International Journal of Computer Science and Network Security
摘要：Increased appearance of robots in both the domestic and professional human life is no wonder today. It needs to improve human-intention recognition capabilities in a robot. Intention recognition is inevitable for effective Human Robot Interaction (HRI). Proactive intention recognition will improve the intuitiveness of HRI. In the presented research work use of reinforcement learning exhibits promising results. Markov Decision Process (MDP) has been used for early intention recognition with the condition of finite state and action space. Different real-time HRI scenarios are modeled using MDPs. A simple algorithm is proposed to identify pseudo destination state(s). Identification of these states is helpful for early intention recognition. An Arduino based robotic arm with a simple webcam is used to perform intention recognition experiments. Use of more sophisticated equipment would enhance the precision level along with the success rate.
