PROBABILISTIC EXTENSIONS TO MODEL BASED PROGRAMMING LANGUAGES

Metamorph is proposing to develop a methodology and tools to apply Probabilistic Programming to System Level Design in our DARPA Phase II SBIR proposal. System design relies significantly on numerical predictions provided by computational models and simulations, however, an objective assessment of confidence in the predictions is a major challenge.Metamorphs proposal is built upon a System Level Design methodology and tools, referred to as the OpenMETA tool suite, developed in prior DARPA funded research and now being productized by Metamorph for transition and commercial applications.This problem of quantifying uncertainty and performing optimization under uncertainty is of significant interest to designers and engineers of complex DoD and Commercial systems. As such, MetaMorph is proposing the following tasks: - Automated development of Bayesian Surrogate Models for System Designs - Automated methods for performing Optimization under Uncertainty - Automated methods for incorporating uncertainty in model libraries - Scaling up Probabilistic Certificate of Correctness (PCC) - (OPTION) Integration of Probabilistic Model Checking for Automated Verification