To search, Click below search items.


All Published Papers Search Service


MDP-led Proactive Intention Recognition for Improved HRI Settings


Muhammad Usman Ashraf, Muhammad Awais, Syed Hasnain Abbas Naqvi, Syed Muhammad Saqlain, Imran Khan and Anwar Ghani


Vol. 19  No. 6  pp. 91-97


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.


Intention Recognition Human-Robot Interaction Markov Decision Process, Reinforcement Learning, Pseudo Destination