The application of statistical reliability theory in the context of Intelligent Environments: a tutorial review

Abstract

Intelligent Environments often require the integration of multi-modal sensing and actuating technologies with high performance real-time computation, including artificial intelligence systems for analysis, learning patterns and reasoning. Such systems may be complex, and involve multiple components. However, in order to make them affordable, Intelligent Environments sometimes require many of their individual components to be low-cost. Nevertheless, in many applications—including safety-critical systems, and systems monitoring the health and well-being of vulnerable individuals, it is essential that these Intelligent Environment systems are reliable, which the issue of affordability must not compromise. If such environments are to find real application and deployment in these types of domain, it is necessary to be able to obtain accurate predictions of how probable any potential failure of the system is in any given timeframe, and of statistical parameters regarding the expected time to the first, or between successive, failures. Such quantities must be kept within what are deemed to be acceptable tolerances if the Intelligent Environment is to be suitable for applications in these critical areas, without requiring excessively high levels of human monitoring and/or intervention. In this paper, an introductory overview of statistical reliability theory is presented. The applicability of this to the context of Intelligent Environments—particularly those involving safety critical or other sensitive issues—is discussed, along with how such reliability modelling can be used to influence the design, implementation and application of an Intelligent Environment.