IS SYSTEM AUTONOMY A QUANTIFIABLE ATTRIBUTE? AN INFORMATION-BASED APPROACH FOR SYSTEM AUTONOMY METRICS PART I: METRICS DEFINITI ON

摘要

In many real world aerospace applications, system Autonomy is the most single significantand meaningful attribute of an Intelligent Autonomous System - IAS. This paper proposesperformance metrics for IAS, which are related to the main dimensions of Autonomy. Metricsare presented and defined. The metrics reflects the most meaningful characteristics of systemAutonomy. The approach taken in this paper defines system or agent autonomy with respectto its self-capability to accomplish its assigned mission goals while operating under uncertainunexpected dynamic environment, uncertain dynamic scenario or self-faulty situations, andwithout or with very little intervention of a human operator or an external agent, throughoutthe mission cycle. The metrics enable to classify the system autonomy level, according to thesensitivity of the system performance to relative changes in autonomy related entities. Themetrics are based on a comprehensive approach. It does not consider the Autonomy level ateach phase of the sense-aware-decide-act process, separately. The proposed metrics arequantitative, relative, and based on concepts from Information theory as well as from Controland Systems theory. Using the System Autonomy Metrics, the system designer can comparebetween competitive designs, architectures or, computational paradigms, as related to thedesired system performance, with respect to Autonomy. These metrics are currently beingused in an on-going research, development and engineering work.