Entropy-based criteria for intelligent autonomous systems

Entropy-based criteria for analysis, evaluation and design of intelligent autonomous systems (IAS) are presented. Based on entropy functions, a global autonomability index (GAI) is defined. Using the GAI, a system can be classified as either completely autonomable, almost completely autonomable, incompletely autonomable, nonautonomable, or nonsupervisable/nonteleoperable. Having the entropy model of an autonomous system, performing an unstructured mission in an uncertain dynamic environment, one can evaluate the performance of the system as a function of the information processing and communication resources, as well as a function of the mission and environment uncertainty, complexity and unexpectedness. Complementary criteria are related to uncertainty convergence rate and information redundancy. The proposed criteria are general and can be used for analysis and evaluation of IAS. The criteria are also applicable as performance-related driving functions in learning, reasoning, dynamic planning, and decision-making processes onboard an IAS