07.01.2013 Views

Lecture Notes in Computer Science 3472

Lecture Notes in Computer Science 3472

Lecture Notes in Computer Science 3472

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

9 Test<strong>in</strong>g Theory for Probabilistic Systems<br />

Verena Wolf<br />

University of Mannheim<br />

Lehrstuhl für Praktische Informatik II<br />

vwolf@pi2.<strong>in</strong>formatik.uni-mannheim.de<br />

9.1 Introduction<br />

The aim of this chapter is to give a survey of test<strong>in</strong>g relations for probabilistic<br />

systems. We summarize the relevant material on probabilistic extensions of the<br />

work of De Nicola and Hennessy [dNH84] who def<strong>in</strong>ed implementation relations<br />

for nondeterm<strong>in</strong>istic processes based on a notion of test<strong>in</strong>g (see also Chapter<br />

5). We ma<strong>in</strong>ly concentrate on the relative expressive power of the different preorders<br />

1 . All presented relations are primarily of theoretical <strong>in</strong>terest and to the<br />

best of our knowledge their usefulness <strong>in</strong> practical applications has not been<br />

shown yet.<br />

Test<strong>in</strong>g can be described as record<strong>in</strong>g the behavior of systems execut<strong>in</strong>g <strong>in</strong> a particularly<br />

designed environment. In the classical sett<strong>in</strong>g a (test<strong>in</strong>g) environment<br />

of a process P is simulated by consider<strong>in</strong>g the parallel composition P�T of P<br />

and a test process T (basically another nondeterm<strong>in</strong>istic process but equipped<br />

with a set of success actions or states). De Nicola and Hennessy def<strong>in</strong>e PmayT<br />

if a success state can (may) be reached by P�T and PmustTif P�T reaches<br />

a success state on every run (execution). Two processes P, Q are related if P<br />

may (must, respectively) T implies Qmay(must, respectively) T for all test<br />

processes T .<br />

In 1990, Ivan Christoff extended the classical test<strong>in</strong>g theory to fully probabilistic<br />

processes that are basically labeled transition systems enriched with probabilistic<br />

<strong>in</strong>formation [Chr90]. He considered the parallel composition P�T of a fully<br />

probabilistic process P and a nondeterm<strong>in</strong>istic process T and analyzed the trace<br />

distribution of (the fully probabilistic result) P�T . Two fully probabilistic processes<br />

P and Q are related if the trace distributions of P�T and Q�T co<strong>in</strong>cide<br />

for all possible test processes T . Additionally, Christoff constructed a characterization<br />

by extended traces which are a denotational model simplify<strong>in</strong>g the<br />

development of algorithms for the computation of the preorder.<br />

Two years later Cleaveland et al. presented a test<strong>in</strong>g approach based on probabilistic<br />

test processes [CSZ92]. They argued that the environment of a fully<br />

probabilistic process P may also be probabilistic and accord<strong>in</strong>gly, they applied<br />

probabilistic test processes T , equipped with success states, and considered the<br />

probability of reach<strong>in</strong>g success <strong>in</strong> P�T . Furthermore, they lifted extended traces<br />

1 We use the term ”preorder” as synonym for ”test<strong>in</strong>g relation” although it might be<br />

the case that the respective relation is not transitive or transitivity has not been<br />

shown.<br />

M. Broy et al. (Eds.): Model-Based Test<strong>in</strong>g of Reactive Systems, LNCS <strong>3472</strong>, pp. 233-275, 2005.<br />

© Spr<strong>in</strong>ger-Verlag Berl<strong>in</strong> Heidelberg 2005

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!