A CSC2108F Conference on Automated Verification

Final Program

The aim of this conference is to bring together students and faculty interested in application of formal methods to specification and verification of software systems.  The conference will consist of a series of presentations by students of CSC2108, Fall'97.  The presentations are based on projects that the students worked on between 12/97-1/98 either on their own or in groups of two.  Projects done by one person will be presented in 30 minutes followed by a 15-minute question period.  Two-person projects will be presented in 45 minutes followed by a 20-minute question period.  The program below gives the schedule of these talks and their abstracts.  Everyone is welcome to attend.  Proceedings of this conference will be published as a DCS technical report.

Program

Abstracts of the Talks

In this project we develop a software controller for a production cell system.  The main objective is to find out if PROMELA/SPIN is an appropriate environment for developping embedded real-time systems.  We describe the production cell hardware model and its requirements.  Subsequently we describe the model of a distributed and concurrent controller for it. Furthermore we mention some required properties we could prove with SPIN based on our model as well as those we could not prove.  We outline a transformation scheme of the PROMELA language to C code.  Finally we describe some experiences and draw conclusions.

Formal methods that are easy to use but with limited flexibility are sometimes criticized as unnecessary.  This presentation will take a detailed  look at a formal method that claims to be user-friendly - UPPAAL.  UPPAAL is a toolbox for modeling, simulating and verifying real-time systems, and components of the UPPAAL toolbox, the modeling language and the verification language will be discussed.  A SCSI bus protocol example is also presented to illustrate its various features. There are discussions regarding advantages and tradeoffs in using UPPAAL. Conclusions are drawn regarding to UPPAAL and "user-friendly" methods in general. 

The Go-Back-N and Selective-Repeat automatic repeat request (ARQ) protocols are widely used for error control in computer-communication networks.  They are nearly optimal for channels characterized by low error rates and small propagation delays, and thus suitable for many classical data transmission applications.  However, these protocols are not efficient for modern non-conventional channels with high error rates and/or large propagation delays, such as mobile and satellite links.  In this talk, we present a family of new protocols, termed Asynchronous Retransmission Go-Back-N ARQ protocols, which offer the potential for better performance in environments characterized by high error rates and/or large propagation delays.  We verify that these protocols are deadlock-free using the Concurrency Workbench.  We also verify that all these new protocols provide the same services to applications as the basic Go-Back-N protocol, and all of them are equivalent to the basic Go-Back-N protocol from the application point of view.

OO lacks a simple model theoretical foundation for definition and discussion, so we need formal

models for the object-oriented programming paradigm.  The major concern of this project is to describe the current state of the art in some of the formalisms in object-oriented analysis and design, their usage, and limitations. 

Some elementary concepts on Equational Theories and Term Rewriting Systems are introduced, followed by a brief review of the main ideas of the Knuth-Bendix completion procedure. Then a formalization of Place/Transition and Coloured Petri Nets as Term Rewriting Systems is proposed, which allows the use of Rewriting Techniques for studying the behaviour of such nets. We verify the relationships existing between Completion, Theorem Proving and Term Simplification and some properties of the Petri Nets, and show how such operations can be used to prove facts about the nets. Particularly, the formalization proposed in this talk leads to an efficient method for determining the movements of the tokens in the net, and it allows a natural way of simplifying nets in order to express more efficiently the movements of tokens when transitions are fired. The theory has been implemented using LP, the Larch Prover, which has been used to derive the results of the examples presented in this talk. 

In this course we looked at tool support for verifying that a design satisfies a specification, with emphasis on problems involving concurrency.  We sometimes hear that formal methods will ensure that we build things right the first time.  There are many obstacles to be overcome before that glorious day, and in the mean time formal methods are being used for much more than verification, including automatic generation of tests and implementations from specifications, software reuse, improved documentation, and design recovery.  This report describes some of the issues in these alternate applications of formal methods, and some of the work being done to resolve these issues.  It also touches on some of the issues for verifying conformance that did not arise during the course, including the choice of what to specify and the meaning of conformance, scale of effort, support for revision, and verifying implementations.

For more information or to request a copy of conference proceedings, send mail to chechik@cs.toronto.edu.  Information about CSC2108 "Automated Verification" is available at www.cs.toronto.edu/~chechik/courses97/csc2108.