Framework of active robot learning

In recent years, cognitive robots have become an attractive research area of Artificial Intelligent (AI). High-order beliefs for cognitive robots regard the robots' thought about their users' intention and preference. The existing approaches to the development of such beliefs through machine learning rely on particular social cues or specifically defined award functions . Therefore, their applications can be limited. This study carried out primary research on active robot learning (ARL) which facilitates a robot to develop high-order beliefs by actively collecting/discovering evidence it needs. The emphasis is on active learning, but not teaching. Hence, social cues and award functions are not necessary. In this study, the framework of ARL was developed. Fuzzy logic was employed in the framework for controlling robot and for identifying high-order beliefs. A simulation environment was set up where a human and a cognitive robot were modelled using MATLAB, and ARL was implemented through simulation. Simulations were also performed in this study where the human and the robot tried to jointly lift a stick and keep the stick level. The simulation results show that under the framework a robot is able to discover the evidence it needs to confirm its user's intention.