engleski

Detecting irregular dynamics of autonomous robotic systems using information entropy

Autonomous systems depend heavily on their information systems. Time and energy consumption, even the probability of reaching the predicted result requires constant information flow & information processing according to underlying information infrastructure, i.e. hardware and software. Because irregular dynamics has finite probability of occurrence it is important to establish measures which, as much as possible, preserve functionality of the whole autonomous system in case of occurrence of several errors. In this article we concentrate on the autonomous robotic systems, starting with the following hypothesis: there are errors in automatic system dynamics which can be traced to larger deviations in their regular dynamics. What is meant by larger deviation, and what are examples of such systems are given in the text. As the method of tracing these errors we utilise information entropy. It is applied onto modelled information flows, simplified representation of realistic information flows taking place within functional autonomous system. We prescribe algorithm for determining the proper information entropy and use it to formulate a criteria of acceptability of deviations in information flow.

information entropy; information flow; automatic robotic system; error

nije evidentirano