Interpreting Fuzzy Linguistic Information by Acquiring Robot's Experience Based on Internal Rehearsal

抄録

This paper proposes a method for interpreting fuzzy linguistic information by acquiring robot's experience on previous movements. The quantitative assessment for a fuzzy linguistic term such as “little” depends on the spatial arrangement of surrounding environment. Robot's experience on previous movements is very useful for understanding such information in a context dependant manner. Therefore, an internal rehearsal system is introduced to acquire the robot's experience on previous movements. Such acquisition enhances the robot's capability to interpret fuzzy linguistic information in fuzzy voice commands (FVCs) according to the current environmental context. The proposed system is used to evaluate the fuzzy linguistic information related to primitive movements of a robot manipulator. Those primitive movements are realized by a behavior evaluation network (BEN) with the guidance of an internal rehearsal system. The end-effector movements and the single joint movements of a robot manipulator are considered as primitive movements. They are activated by fuzzy voice motion commands (FVMCs) and fuzzy voice joint commands (FVJCs) respectively. The proposed idea is demonstrated with a PA-10 robot manipulator by navigating the robot manipulator in the user's working space. The use of the behavior evaluation network and the internal rehearsal system introduced a better way of assessing fuzzy linguistic information by acquiring the robot's experience on the corresponding environment.