Automatic generation of dependency models using autonomous intelligent agents

Dependency models are the basis of several important products for testability analysis, diagnosability analysis, and generation of optimal fault trees including generation of dynamic test strategies based on current parameters and available resources. The AI-ESTATE Committee of the IEEE SCC20 is working on a standard representation of dependency models (IEEE 1232.1) and once that is formalized, the resulting increased portability and re-usability of models will make dependency models even more important. The problem with dependency models is that they are difficult to generate. Dependency models generated from netlist information alone are close to useless. Dependency models can also be generated via a circuit simulator, but these run very slowly and for complex circuits, this is not feasible. Intelligent Automation, Inc. (IAI), together with the U.S. Army Communications and Electronics Command, is developing a new approach to automatic generation of dependency models. The approach is based on use of “intelligent agents”. Intelligent agents may also be useful for other aspects of the problem of testing and diagnosing complex electronic circuits, and our success in the automatic generation of dependency models has led us to begin studying other applications of agents in this domain. There are two key benefits of implementing software using the paradigm of autonomous agents. First, agents are independent and asynchronous, which means they can run on any number of computers which might be available as long as the computers are networked together. The second benefit is ease of implementation. The work reported here was done as an experiment in the use of autonomous agents and not with the goal of implementing a usable system