Zeigler’s previous books have concentrated on building a much-needed theoretical foundation for simulation methodology. His latest volume, however, describes the DEVS-Scheme modeling and simulation environment, which represents the fruits of his and his students’ labors in implementing his ideas on multifaceted modeling. This book should be regarded as more than a description of a particular simulation software system. In providing a detailed illustration of one particular implementation of his own theories, Zeigler tackles many issues relevant to present-day simulations. Among the topics central to Zeigler’s system are the incorporation of artificial intelligence techniques into simulation environments; the need for a hierarchical, modular approach to modeling; and object-orientation as a natural paradigm for expressing and implementing the resulting models.

Before going into the details of DEVS-Scheme, the author lays the foundations for his work, discussing knowledge representation schemes and formalisms that are relevant to simulation and simulation environments. This introduction is followed by a chapter entitled “Basics,” which contains a section on the concepts of object-oriented programming. I did not like this section, in part because Zeigler says that “object-oriented programming is a paradigm in which a software system is decomposed into subsystems based on objects.” The view usually held is that object-oriented design, programming, and simulation are based on composition, not decomposition. Objects (or sub-models) are composed to form larger objects (models). I did not like his definition of abstraction, which would be a good definition for encapsulation (which does help to make abstraction possible) but not for abstraction itself. The next section introduces the system entity structure/model base (SES/MB), which is important for supporting knowledge-based environments. The SES represents and controls the composition of models from components in the model base. The chapter closes with a large example of a general SES for a life support system such as might be required on the moon. The structure is then pruned in order to develop a specific simulation model.

Subsequent chapters introduce, illustrate by examples, and elaborate the details of the DEVS-Scheme environment.  Zeigler  and his system primarily address discrete event simulation; continuous system simulation receives only a passing mention. (Work has been done, however, on incorporating continuous formalisms within the DEVS-Scheme environment.) One nice feature of the text is the introduction of a model of a single-processor computer system, which is extended, as the details of the environment are unfurled, to include multiple processors and system synchronization. Strange though it may seem in a discussion of a discrete-event system, none of Zeigler’s examples uses queues, because he says that queues can indicate inadequate process coordination and thereby incur a costly overhead. This gives the erroneous impression that all queues are inherently expensive and should be avoided. In my experience, “more sophisticated co-ordination schemes,” as Zeigler calls them, do not always work as well as intended, if at all. What is more, many real-world systems do contain queues, and it is to be assumed that they could be represented in DEVS-Scheme, even if Zeigler does not show how.

Zeigler devotes several chapters to endomorphism, that is, a model that includes a model of a model. If one is modeling an entity that has intelligence, such as a human decision maker, it may be necessary to represent the model that the intelligent agent uses in order to make decisions. One of the  interesting  issues discussed in this context is the process of deriving different models of the same system through abstraction. Zeigler includes some interesting and relevant quotes from Marvin Minsky’s Society of mind [1], but  Zeigler’s  billing Minsky’s book as “the most comprehensive attempt to understand mind and brain” is erroneous.

Although Zeigler claims that DEVS-Scheme is suitable at least as a basis for a distributed simulation system, he does not fully elaborate the grounds for this claim. In DEVS-Scheme, simulations are controlled by virtual processors, which would seem to provide an excellent basis for a distributed mapping. It is not clear what synchronization facilities virtual processors have to support distributed execution, however.

The penultimate chapter puts DEVS-Scheme in the context of the work of other people in the field. I would have found this discussion more useful as an introduction rather than as a postscript. In evaluating what is to come, I prefer to know from the outset what the author’s position is and where it fits in “the larger scheme of things,” as Zeigler rightly calls it. I also feel that this review is rather selective. It omits other work that, while perhaps not as ambitious as that of Zeigler, is nonetheless apposite.

The proofreading and copy editing passed by a number of errors, including minor errors in figures, an incorrect section heading, and several sentences that would defy any parser. The index has been sorted in ASCII sequence, so co-ordinators is placed between closure and coherence. I wish publishers would remember that books are written for humans.

Anybody who is interested in any of the issues discussed above should read this book, for it describes a significant step on the road to achieving knowledge-based simulation. It matters not if you agree or disagree with Zeigler’s theories and implementation; the fact that he has produced a working system affords a focus about which to form your own opinions and clarify your own ideas.