Bayes parameter identification for optimal control of robots

Abstract

Deals with parameter identification using the fuzzy Bayes decision rule, which makes it possible to take into account the differing consequences of positive and negative estimation errors. Nonsymmetrical loss functions, in piecewise linear and piecewise quadratic forms are considered. The calculation procedures have been based on the statistical kernel estimators methodology. An example application for the time-optimal and quadratic optimal controls for a fundamental mechanical system is described. The concept proposed can easily be transferred to sophisticated tasks of contemporary robot control.