simulation@uflorida.cis.ufl.edu (Moderator: Paul Fishwick) (11/22/89)
Volume: 12, Issue: 8, Wed Nov 22 09:55:25 EST 1989 +----------------+ | TODAY'S TOPICS | +----------------+ (1) High Level Hardware Design (2) Petri Net Simulation (3) Overview of DMOD * Moderator: Paul Fishwick, Univ. of Florida * Send topical mail to: simulation@bikini.cis.ufl.edu OR post to comp.simulation via USENET * Archives available via FTP to bikini.cis.ufl.edu, login as 'ftp', use your last name as the password, change directory to pub/simdigest. * Simulation Tools available by doing above and changing the directory to pub/simdigest/tools. ----------------------------------------------------------------------------- From: "Wolfgang Mueller" <cadlab!wolfgang@uunet.UU.NET> Subject: HIGH-LEVEL DESIGN To: fishwick@bikini.cis.ufl.edu Date: Mon, 20 Nov 89 11:17:08 MET DST X-Mailer: Elm [version 2.1 PL1] Hello, I've read your reference request in the net in comp.simulation about high level design simulation. I have been looking for a long time for a reference list about this topic and I would be glad to receive one. Maybe you can tell me whether a summary article about high-level design (tools & environments) will be published or maybe you can send a summary of your received references to the net. kind regards wolfgang mueller ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ | Wolfgang Mueller | Tel. : (+49) (+) 5251-284128 | | CADLAB | Fax : (+49) (+) 5251-284140 | | Bahnhofstr. 32 | | | 4790 Paderborn | E-Mail: wolfgang@cadlab.uucp | | F.R.G. | wolfgang@cadlab.cadlab.de | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ------------------------------ From: theseas!illusion@relay.EU.net (Charalambos N. ATHANASSIOU) Date: 21 Nov 89 21:44:25 GMT To: comp-simulation@csi.forth.gr Subject: Submission for comp-simulation Responding-System: theseas Path: theseas!illusion From: illusion@theseas (Charalambos N. ATHANASSIOU) Newsgroups: comp.theory,comp.parallel,comp.simulation Subject: Petri-Net information enquiry Keywords: Petri-Net,simulation,algorithms Date: 21 Nov 89 21:44:24 GMT Organization: National Technical University of Athens Sorry if I'm posting in the wrong newsgroups, but I didn't know of any specific ones. Does anyone has any information and/or pointers, e-mail addresses on the subject of Petri-Net simulation/analysis? In particular I seek references about tools, text books, algorithms for P-N analysis. [[There currently is one tool for timed nets in the tools directory -- see the header of this digest issue. Also, Andreas Nowatzyk of CMU has designed a "general stochastic petri net simulator" which I will try to insert in the tools library if there is interest in the readership. I will need to get in touch with him to check on availability for mass distribution -PAF]] I'm working on my graduation thesis and any help would be appreciated. Please reply to me with e-mail and I'll post a summary to each newsgroup if enough interest is generated. -- Charalambos N. ATHANASSIOU E-mail: c/o Prof. Spyros Tzafestas UUCP: mcvax!ariadne!theseas!illusion Lab. of Robotics Control InterNet: illusion%theseas.uucp@uunet.UU.NET and Expert Systems ------------------------------ To: simulation@bikini.cis.ufl.edu Cc: narain%pluto@rand.org Subject: Overview of DMOD Date: Tue, 21 Nov 89 17:29:50 PST From: narain%pluto@rand.org This is a highly belated response to Steve Glicker's suggestion that I post an outline of our new simulation formalism, DMOD, to this digest. The main reason for the delay is that our own understanding of DMOD has been evolving. In case the following is too long, a paper on DMOD "A logic for simulating discontinuous systems" will be presented at the upcoming WSC. OVERVIEW OF DMOD ================ DMOD has been developed under the RAND Advanced Simulation Language project led jointly by Jeff Rothenberg and myself. The purpose of the project is to develop techniques which would drastically reduce the complexity of building models of dynamic systems and of reasoning about them. The systems of interest can be discrete, **continuous** or combinations of the two. DMOD is a first step in our attempts to achieve our goals. It is based upon the following fundamental assumption: If all event occurrences in a system till time T are known, the state of the system can be computed at any point of time till T. Thus, simulation is regarded as computation of event occurrences. It is proposed that a convenient way of specifying event occurrences is via the causality relation. Intuitively, "E causes F" means "E is responsible for or brings about F". Hence, if E occurs, the occurrence of F can be inferred. Now, a model of a system can be regarded as a definition of the causality relation for it, and simulation as the inference of event occurrences from it. As we have well developed intuitions about causation, such a definition can be quite easy to provide. A ready-made implementation of this scheme can be obtained in relational languages such as Prolog. However, rules defining causation frequently refer to the past and the future of causing events. For example, we have: An event of a customer requesting credit at T causes an event of credit being granted at T provided the customer has never defaulted before T. An event of dialing a phone number at T causes an event of that phone to ring after T+10 seconds provided the dialer does not hang up in between. Unfortunately, it is beyond the capability of Prolog and many other automatic deduction systems to infer event occurrences from such rules. A central contribution of DMOD is a method of drastically alleviating the difficulty of reasoning with such rules. It is based upon an alternative view of causality in which we regard it, not as a binary relation, but as a ternary relation between two events and a context of a temporally ordered sequence of events. Rules of the above form can be reexpressed, using the new view, in such a way that they are very easy to reason with. A new algorithm is developed for inferring event occurrences from such rules. The scheme can be implemented conveniently in Prolog. DMOD exhibits the following advantages: (1) Ease of model development ----------------------------- As mentioned above, causality is a natural relation. It is reasonable to assume that every event, except the first one, has a cause. Thus, a definition of causality can be easy to provide. Also, models can possess an intuitive character. Events are said to occur when interesting conditions become true. This allows continuous systems to be modeled within DMOD. For example, a condition can be that the temperature of a reactor has reached critical point from below. Whenever it becomes true, an event can be said to occur and we can write down rules specifying what events could cause it. (2) Useful forms of analyses ---------------------------- DMOD treats histories as first-class citizens, i.e., as manipulable objects. This makes practical very useful types of analyses e.g., simulation backup, tracing of causality chains, or parameter instrumentation. (3) Formalization and simplification of discrete-event simulation ----------------------------------------------------------------- DMOD is proposed as a formalization of the ideas behind the widely used event-scheduling view of the discrete-event technique. Discrete-event models employ operations of scheduling and unscheduling of events upon an event queue. These can be viewed as procedures for inferring event occurrences from abstract causality rules in the mind of the modeler. For example if E schedules F in the future, but if G occurs in between, it unschedules F, we can formulate the following causality rule: E causes F if G does not occur in between E and F. In contrast, a DMOD model directly consists of causality rules, not of a procedure for inferring event occurrences. Algorithms for event occurrences are separated from the DMOD model and are, in fact, invisible to the modeler. This view of discrete-event simulation reveals that devices of event queues and scheduling and unscheduling are logically unnecessary. In other words, it is possible to infer event occurrences from causality rules without making any use of these devices. This yields considerable simplification in the reasoning about DMOD models. (4) Basis for temporal reasoning -------------------------------- The declarative nature of DMOD, and the abstractions employed in it, yield new methods for proving temporal properties of dynamic systems. The methods can also be employed for continuous systems. Most contemporary temporal logics seem to be intended only for discrete systems. Furthermore, proofs are carried out within first-order logic using simple axioms about causality and the model itself. No new logic with a different syntax or semantics is introduced. Formal methods of reasoning about models are, of course, essential for designing advanced algorithms such as for intelligent debugging, model composition, lazy simulation, real-time planning or answering complex questions about the model. CURRENT STATUS ============== A simple implementation of DMOD has been developed in Quintus Prolog. We have demonstrated its usefulness by building a fairly realistic logistics model in it. It is still a research vehicle, however, and is not ready for release. Our current work consists of developing advanced algorithms for problems mentioned in (4) above. Comments or criticisms are invited. Sanjai Narain RAND Corporation ------------------------------ END OF SIMULATION DIGEST ************************