Analyzing Dynamic Aspects of AxC Systems by Means of Statistical Model Checking

Many researchers shown that approximate circuits are able to provide a new perspective on the development of electronic systems. Mostly, they tried to find an optimal trade-off between the approximation error and resource savings for predefined applications. However, they used to concentrate mainly on design aspects regarding relaxed functional requirements, but neglected aspects like timing, sequential/asynchronous nature of circuits, uncertainty due to process/parameter variations, excessively high operating frequencies or low voltages. This paper aims to take a step ahead by moving towards the verification of dynamic properties of systems based on approximate circuits, with a focus on sequential/asynchronous circuits and uncertainty. First, the paper presents our approach to modeling approximate systems by means of stochastic hybrid timed automata. Then, it shows the principle/advantage of verifying properties of modeled systems by the so-called statistical model checking technique. Further, it presents a framework that takes at its input the model of an accurate system, its timing and other requirements and expected properties, information about basic building blocks and acceptable cost/quality trade-off to produce an approximated system that meets the requirements maximally and satisfies the properties. Finally, the paper evaluates our approach and outlines future research perspectives.