Abstract
Component based modeling has been a research hotpot in the area of Modeling and Simulation for a long time. In order to increase the reusability and interoperability of models, this paper put forward a component based modeling framework called BOM Modeling Framework (BMF) based on BOM specification. The structure, principles, formalisms and behaviors of BMF were presented in detail first. The description and framework of Atomic BOM model and coupled BOM model in BMF were studied in succession. At last, the developing process of BMF models was introduced. The prototype developed showed that BMF could not only promote the reusability and interoperability of models but also improve the efficiency and easy the process of M&S.
Introduction
In the field of Modeling and Simulation (M&S), the High Level Architecture (HLA) has long been a standard for distributed simulation. However, a persistent challenge remains: "federates" (simulation components) are often tightly coupled to a specific Federation Object Model (FOM). This tight coupling makes it difficult to reuse models across different projects, as developers must often rewrite interfaces and data mappings when moving a component to a new environment. The BOM Modeling Framework (BMF) addresses this "reusability headache" by decoupling simulation logic from the underlying execution engine, allowing for a modular, component-based approach to simulation design.
The BOM Modeling Framework (BMF)
The BMF is built upon the SISO Base Object Model (BOM) specification and utilizes several key architectural strategies to ensure model independence and interoperability.
The KM/CM Split
The core innovation of BMF is the separation of a model into two distinct parts: the Kernel Model (KM) and the Connected Model (CM). The KM contains the pure simulation logic—the "what" of the model—and remains entirely agnostic of the simulation engine. The CM acts as a middleware or proxy—the "how"—handling all communication with the engine and other models.
Layered Architecture and Execution
BMF organizes simulations into a four-layer hierarchy: Atomic Model, Coupled Model, Federate, and Federation. This structure is supported by the XSRFrame (Extensible Simulation Running Framework) and the Run-time Infrastructure (RTI) for distributed synchronization. The framework employs the Parallel Discrete Event System Specification (DEVS) formalism to mathematically define state transitions and event handling.
To streamline development, the authors introduced automated tooling, including a ModelEditor for XML-based model definition and a Connect Model Generator that automatically produces the CM code, significantly reducing manual effort and potential errors.
Key Insights and Design Principles
- Double Abstraction: BMF hides implementation details between models, enabling "plug-and-play" interoperability. Models can invoke actions on others without knowledge of their internal code.
- Publish/Subscribe Mechanism: Moving away from direct-binding, BMF uses a publish/subscribe model. Components announce the data they produce or require, allowing the simulation topology to adapt dynamically.
- XML-Centric and Modular Packaging: All model descriptions and mappings are stored in XML, creating a searchable repository. Models are packaged as Dynamic Link Libraries (DLLs), allowing them to be loaded into simulations without requiring a full system recompile.
Results and Future Work
A prototype of the BMF was developed to validate these concepts. Qualitative results indicate significant improvements in development efficiency and model portability. A key success was the demonstrated reuse of the same Atomic Model (e.g., "Model A") across entirely different federates, as seen in the architectural overview. By auto-generating the Connected Model, researchers were able to focus their efforts solely on the simulation logic within the Kernel Model.
Despite these successes, the authors identified several areas for future development:
- Composition Depth: Expanding the framework to support multi-level, nested hierarchies beyond the current two-level limit.
- Runtime Flexibility: Implementing mechanisms to change state variables and parameters dynamically during execution.
- Distributed Performance: Developing a customized parallel simulation engine to overcome performance bottlenecks and data conversion overhead associated with standard RTI in high-performance computing environments.
Summary updated