Large-Scale Integration of Heterogeneous Simulations

Abstract

Systems-of-Systems (SoS) are composed of several interacting and interdependent systems that necessitate the integration of complex, heterogeneous models that represent the ensemble from different points of view, such as organizational workflows, cyber infrastructure, and various engineering or physical domains. These models are complex and require different dynamic simulators to compute their behavior over time. Thus, evaluation of SoS as-a-whole necessitates integration of these heterogeneous simulators. This is highly challenging because it requires integrating both the heterogeneous system models with different semantics and concepts from different system domains (physical, computational, or human), and the heterogeneous system simulators that use different time-stepping and event handling methods. Further, real-world SoS simulation and experimentation requires a comprehensive framework for integration modeling, efficient model and system composition, parametric experiments, run-time deployment, simulation control, scenario-based experimentation, and system analysis.

This dissertation presents a model-based integration approach for integrating large-scale heterogeneous simulations. The approach is illustrated by developing a generic simulation integration and experimentation framework called the Command and Control Wind Tunnel (C2WT). It allows modeling systems with their interdependencies as well as connecting and relating the corresponding heterogeneous simulators in a logically and temporally coherent manner. Its generalizable methods and tools enable rapid synthesis of industry standards based integrated simulations. For real-world integrated simulation experiments, several novel techniques are presented such as mapping methods for integrating legacy components that cannot directly interface with SoS-level data models, a generic cyber communication network simulation component that can be reused for different SoSs, a reusable cyber-attack library for evaluating SoS’ security and resilience against cyber threats, and modeling and orchestration of alternative what-if scenarios for SoS evaluations. Further, for efficient simulation of complex dynamical models that exhibit different rate dynamics in different parts, a partitioning method is developed to split them into different sampling rate groups. In addition, a novel approach is presented for ontology based model composition. In-depth case studies are also provided to demonstrate the effectiveness of the overall integration approach.