PROBABILISTIC EXTENSIONS TO MODEL BASED PROGRAMMING LANGUAGES

This project will demonstrate the feasibility of extending the probabilistic design verification capabilities of the OpenMETA tool suite with probabilistic programming methods. These extensions will enable correct-by-construction design methodologies in the presence of component, system, environment, and epistemic uncertainties.The OpenMETA model- and component-based design tool suite for the manufacturing-aware design of complex Cyber-Physical Systems (CPS) uses a deterministic model for dynamic simulations, using a causal component model connections, with lumped parameter models represented in Modelica.The current, limited PCC approach allows stochastic modulation of parameters in a Monte Carlo simulation. Thisproject will design and test mechanisms for introducing Probabilistic Dynamics Modeling directly in Modelica, a significant capability enhancement of the Modelica . These extensions will add probabilistic variablesparameters, signals, input/outputs to the model that could vary within a simulation run, and could be used to accurately incorporate the uncertainty in the dynamics models.The current PCC is insufficient for capturingtime-dependent, non-invariant signals. The proposal will add a probabilistic dynamics model to Modelica to address these limitations. Furthermore, the project will investigate mechanisms for addressing Inverse Problems, using inferencing techniques (Bayesian, MCMC), and sampling reduction approaches