Abstract
This paper provides a comprehensive architectural overview of the Modeling and Simulation (M&S) landscape, establishing a unified "Big Picture" framework. It bridges the gap between theoretical formalisms and practical engineering by introducing rigorous taxonomies for simulation data and a "Model-as-a-Service" (MaaS) framework to address challenges in model reuse and credibility.
Introduction
The field of Modeling and Simulation (M&S) has historically been fragmented. Professionals from diverse backgrounds—such as electrical engineering, mathematics, and social sciences—often use the term "model" to describe vastly different concepts and goals. This lack of a unified Body of Knowledge (BoK) has led to several critical issues:
- The "Black Box" Problem: Legacy models are frequently undocumented and tied to specific software, hindering reuse.
- Credibility Gaps: Without standardized lifecycles, verifying if a simulation accurately represents reality is difficult.
- Data Ambiguity: The absence of a standard vocabulary for simulation data leads to confusion.
To address these, the authors propose a structured, engineering-centric approach to the entire simulation lifecycle, aiming for more credible, reusable, and interoperable systems.
Methodological Framework
The paper synthesizes several high-level methodologies to create a cohesive framework for M&S:
Taxonomical Classification
The authors provide exhaustive categorizations for reality/model dichotomies and simulation perceptions. This includes a rigorous taxonomy for over 40 distinct types of simulation data (e.g., livewired vs. hardwired, notional vs. actual).
Model Life Cycle Engineering (MLE)
A systematic 6-phase process—Definition, Design, Construction, VV&A (Verification, Validation, and Accreditation), Application, and Evolution/Maintenance—is introduced to ensure model credibility throughout its existence.
System Entity Structure (SES) and Statistical Heuristics
The SES methodology is used for organizing families of models in repositories to support automated composition. Furthermore, the paper reviews statistical techniques for "Transient Elimination," such as "Batch Means" and "Initial Data Deletion," to manage non-stationary "warm-up" noise in steady-state simulations.
Key Insights and Ideas
- The Cultural Lens: Modeling is viewed as fundamentally cultural. For an artist, reality is the inspiration (model); for a simulationist, the model is the representation (reality). Bridging these views is vital for interdisciplinary work.
- Model Curation: The authors argue for "curation" rather than just storage. Models should be annotated with theoretical foundations, like Experimental Frames, to facilitate automated retrieval and integration.
- Model as a Service (MaaS): A shift toward "servitization" is envisioned, where models and execution engines are cloud-encapsulated, allowing for autonomous evolution and easier sharing.
Key Results
The primary outcome of this work is the establishment of standardized frameworks:
- Unified Data Vocabulary: A "lingua franca" for M&S practitioners defining dozens of data types.
- MLE Framework: A visible pipeline for model management that elevates M&S to a full-lifecycle engineering discipline.
- Formalism Integration: A mapping of how different system specifications (DESS, DTSS, and DEVS) can coexist within a single "multimodeling" umbrella.
Limitations and Future Work
Despite these advancements, several hurdles remain:
- Scalability: Techniques for establishing and searching massive, multi-disciplinary model libraries are still in their infancy.
- Computational Intractability: Automated model composition remains a significant theoretical and technical challenge.
- Quantitative VV&A: There is a pressing need for more formalized, quantitative methods for Verification, Validation, and Accreditation, as current research still leans heavily on qualitative analysis.
Summary updated