Performance improvement of intelligent machines through feedback

Abstract

This paper introduces an algorithm for performance improvement of intelligent machines based on a cost function recursively estimated from feedback. The interfaces between the three levels of the hierarchical intelligent controller (HIC) for the intelligent machine are modeled by a 2-stage hierarchical learning stochastic automaton (HLSA). The cost function used by the HLSA combines measures of reliability and computational cost, defined in conjunction. Novel contributions of the paper include an original hierarchical reinforcement learning scheme and a new cost function for intelligent machines. Results of simulations show the application of the methodology to an intelligent robotic system.