Bernard P. Zeigler

Activity-Based Credit Assignment Heuristic for Simulation-Based Stochastic Search in a Hierarchical Model Base of Systems

Synthesis of systems constitutes a vast class of problems. Although machine learning techniques operate at the functional level, little attention has been paid to system synthesis using a hierarchical model base. This paper develops an original approach for automatically rating component systems and composing them according to the experimental frames in which they are placed. Components are assigned credit by correlating measures of their participation (activity) in simulation runs with run outcomes. These ratings are employed to bias component selection in subsequent compositions.

System Entity Structure Basics

In this chapter, you will see how the System Entity Structure (SES) can help you construct models for Systems of Systems. In fact, we will use the SES to better understand the process for constructing such models. Modeling and Simulation (M&S) refers to a set of activities that are undertaken for a variety of reasons. We get a sense of the activities involved in M&S from a bird’s eye perspective before we dive down into DEVS-based tools for simulation model construction, the focus of this book. We formulate some of the basic activities in Modeling and Simulation as a process or sequence of steps that can be represented with a System Entity Structure. Besides conveying some familiarity with M&S activities, we use the example to discuss two unique features of the SES, decomposition and coupling.

DEVS Integrated Development Environments

This book is divided into three parts. In the first part, we discuss basic DEVS and SES concepts and tools to support working with these concepts in the context of an actual modeling and simulation environment, called MS4 Modeling Environment (MS4 Me™). Then in Part II, we discuss more advanced concepts that such tools can support, and in Part III, we discuss some actual applications that throw light on the kinds of System of Systems problems that can be addressed with such concepts and tools.

Aspects and Multi-aspects

This chapter starts with a discussion of how different aspects can be associated with the same entity and how this allows you to decompose a system in different ways. This leads to a consideration of the concept of multi-aspect which provides a uniform way to associate an unlimited number of related aspects with the same entity. Pruning a multi-aspect involves setting its multiplicity and restructuring it into an ordinary aspect with the specified number of components. We show how pruning of multi-aspects effectively open up a large space of simulation models with an unbounded variety of possibilities for coupling their components. Unfortunately, unless properly managed, this variety can also entail enormous amounts of detailed data entry which can be tedious and error prone. This leads to development of a uniform coupling rule which separates node-to-node network connectivity (specified by a directed graph) and port-to-port coupling which is forced to be uniform across all network connections. Some commonly employed schemes such as cyclic, cellular, and tree compositions have well defined digraphs with uniform couplings so they fit this mold.

Languages for Constructing DEVS Models

This chapter first provides a higher-level perspective on the approach that MS4 Me™ takes to computational support for constructing DEVS models for virtual build and test. After describing this approach, we expand our view to examine whether Unified Modeling Language (UML) can provide a more expressive framework for DEVS specification. For completeness, we also look at how UML can serve as a target for implementation of DEVS models.

Automated and Rule-Based Pruning

The main features of the System Entity Structure, its specializations and aspects, as well as pruning and model generation have now been introduced. Such concepts provide a wealth and variety of potential hierarchical structures with which to tackle complex Systems of Systems problems. However, the rapidly growing combinatorial spaces that are set up by specialization and aspect selections can outstrip human capacity to do manual pruning. Accordingly, this chapter discusses automated pruning—concepts and tools for pruning that can reduce, and sometimes, eliminate, the manual pruning that is otherwise required. Enumerative pruning entirely eliminates manual pruning entirely but is restricted to small enough solution spaces. Random pruning samples from a large solution space to give a statistical picture of the space. Context free and context sensitive selection rules provide the ability to constrain the solution space to combinations that are more likely to meet your requirements.

Managing Inheritance in Pruning

Pruning a System Entity Structure involves selecting aspects from entities as well as entities from specializations. In particular, selecting an entity (the child) from a specialization under another entity (the parent) results in a combination of child and parent that can inherit some of the properties of the parent or child. We show how MS4 Me™ allows flexibility in how you want this inheritance to be carried out. The inheritance specifications are added to the pruning script and control the hierarchical model generated by transforming the pruning entity structure.

DEVS Natural Language Models and Elaborations

MS4 Me™ provides a primary means of creating simulation models through transformation of Finite Deterministic DEVS (FDDEVS) specifications. This chapter provides an understanding first, of the FDDEVS models, and second that how these files get transformed into DEVS atomic models expressed in Java. We then show how you can enhance FDDEVS models to enable to them to automatically generate DEVS atomic models in Java that have full capability to express messages and states. We also show how hierarchical models can be created using the Sequence Designer and then enhanced using the FDDEVS elaboration process.

Modeling and Simulation of Living Systems as Systems of Systems

In this chapter, we show that the system theoretic basis of the DEVS formalism matches the systemic point of view adopted in the living sciences field. Two examples, one in animal epidemiology and the other in plant growth modeling, illustrate different characteristics of DEVS and its extensions. We show how these multi-formalistic abilities of DEVS Modeling Environments are very promising to help answer critical issues regarding natural risk management and poverty reduction. We show how DEVS can serve as a universal formalism for living dynamical system modeling and simulation. However, DEVS is an abstract formalism and can be hard to manage when the modeling effort has to focus on the application domain. So the Virtual Laboratory Environment (VLE) provides an environment where DEVS is used at the simulation level but where the modeling level is composed by a set of specialized modeling components, where components are represented by appropriate formalisms. In this way, the modeler can design the model using the most suitable formalism, or coupling several ones, without any knowledge of DEVS. VLE supplies the mappings into DEVS of the main formalisms used in living system modeling and simulation. We show how the tackling of urgent issues, such as the economical and ecological crises, diversity erosion, or poverty, will be based on creative compositions of shared representations among multiple discipline-based experts. These shared models will embed heterogeneous knowledge elements at different scales, i.e., in heterogeneous formalisms. For these reasons, simulation models for living systems, viewed as systems of systems, are best formalized with DEVS and its extensions.

Flexible Modeling Support Environments

In this chapter, we discuss a Modeling Support Environment (MSE) whose goal is to provide the flexibility to adapt its workflows, tools, and models, to diverse stakeholders. We outline the unique features of the MSE that support its use by a wide spectrum of potential users and developers of a system of fractionated spacecraft. These features include identification of user types to enable routing the user through relevant processing stages, automated generation of model artifacts adapted to selected pathways, conditioning of the solutions space to increase the opportunities to find suitable fractionated architectures, flexible simulation services, and consistent configuration across multiple abstraction models. and semantics-based orchestration of service oriented architecture. The approach taken in the design and development of the MSE is based on fundamental principles that have application much beyond spacecraft fractionated systems. This generic quality of the MSE concept suggests the applicability of DEVS Modeling Environments to virtual build and test of today’s system of systems.