Adaptive selection of intelligent processing modules and its applications

Abstract

In this paper we study the problem of applicationbased Human-Robot Interaction (HRI). We introduce a problem called The Human State Problem (HSP) and we propose a robotic architecture that partially solves this problem. In the HSP the goal is to keep a user that interacts with a robotic application in a desired state; in most cases this state is happy or satisfied. The robotic application uses real world feedback to reconfigure its behavior. The behavior is generated by a selection mechanism that adaptively selects computational resources that are then used for the processing of the current input-to-output mapping. The computational resources are selected from a pool of available intelligent processing resources that represents all available computational capacity of the robotic application. The main problem is in the fact that the robotic application receives only indirect and partial human feedback. Such feedback is not sufficient for the robot to easily predict or decide what actions are the most appropriate.