A novel software architecture for computer-aided analysis of circuits with uncertain parameters

In recent years interest has grown in the possibility of applying novel software technologies (such as those from artificial intelligence) to the field of scientific calculus. There is an increasing need of such approaches due to the requests for integration-cooperation features in modern applications, whose simulation tasks are asked to go far beyond the rigid application of presumed general purpose numerical methods. In this paper we explore the use of software agents for circuit-CAD (CCAD) with uncertainty handling capabilities. In such a field, solving some hard problems like tolerance and sensitivity analysis (TSA) for strongly nonlinear circuits in presence of large parameter uncertainties, benefits much more from a clever preliminary choice of the best problem-tuned numerical method rather than from a blind number crunching approach. Software agents seem to meet these constraints while introducing a high degree of flexibility, which enables the implementation of dynamic adaptive resolution schemes for given CCAD tasks. We present an example of use of the proposed architecture utilizing Java to specify agents as active objects. Some applications of the new strategy to TSA circuit-related problems are presented