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

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

240 Verena Wolf<br />

1<br />

2<br />

sP<br />

a b<br />

µ λ<br />

1<br />

2<br />

2<br />

3<br />

1<br />

3<br />

u1 u2 u3<br />

Fig. 9.2. The probabilistic process P.<br />

the environment by resolv<strong>in</strong>g nondeterm<strong>in</strong>ism. So one probabilistic process P<br />

describes a set of fully probabilistic processes (here called resolutions of P). We<br />

show how to f<strong>in</strong>d a way of splitt<strong>in</strong>g P <strong>in</strong>to several resolutions, i.e. how to remove<br />

nondeterm<strong>in</strong>ism and obta<strong>in</strong> a set of fully probabilistic processes (resolutions)<br />

from P. A more detailed comparison between several models for probabilistic<br />

systems can be found <strong>in</strong> the work of Glabbeek, Smolka and Steffen [vGSS95].<br />

Def<strong>in</strong>ition 9.3. A probabilistic process is a tuple P =(SP, →P, sP), where<br />

• SP is a countable set of states,<br />

•→P⊆SP × Actτ × Distr(SP) is a transition relation,<br />

• sP ∈ SP is an <strong>in</strong>itial state.<br />

⊓⊔<br />

We write s a −→P µ for a transition (s, a,µ) ∈→P and def<strong>in</strong>e PP as the set of<br />

all probabilistic processes. A state s can have several nondeterm<strong>in</strong>istic alternatives<br />

for the next transition. The dest<strong>in</strong>ation of a transition s a −→ µ is chosen<br />

probabilistically with regard to the distribution µ. For a transition s a −→ µ the<br />

probability µ(s ′ )forsomes ′ ∈ supp(µ) can be seen as the conditional probability<br />

of the target state s ′ given that transition s a −→ µ is chosen. This is motivated<br />

by the idea that the external environment decides which action is performed<br />

whereas the probabilistic choice determ<strong>in</strong><strong>in</strong>g the next target state is resolved<br />

by the process itself. Of course, the external user cannot choose between two<br />

transitions s a −→P µ and s a −→P µ ′ , µ �= µ ′ . We therefore have a k<strong>in</strong>d of ”true”<br />

nondeterm<strong>in</strong>ism <strong>in</strong> addition. Later on we will remove nondeterm<strong>in</strong>ism by add<strong>in</strong>g<br />

weights to every transition such that analyz<strong>in</strong>g the process becomes less difficult.<br />

Example. Figure 9.2 shows P =(SP, →P, sP) ∈ PP with SP = {sP, u1, u2, u3},<br />

→P= {(sP, a,µ), (sP, b,λ)}, µ(u1) =µ(u2) = 1<br />

2 , λ(u2) = 2<br />

3 and λ(u3) = 1<br />

3 .<br />

States are drawn as circles and distributions as boxes. Transitions s a −→P µ are<br />

drawn as solid arrows and probabilistic choices are drawn as dashed arrows.<br />

We can state that the probability of reach<strong>in</strong>g u1 lies <strong>in</strong> the <strong>in</strong>terval [0, 1<br />

2<br />

a higher probability than 1<br />

2<br />

] because<br />

is not possible even if the environment schedules the

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

Saved successfully!

Ooh no, something went wrong!