24

Petri Nets

Author: Robert A. McGuigan, Department of Mathematics, Westfield State

College.

Prerequisites: The prerequisites for this chapter are graphs and digraphs.
See Sections 9.1, 9.2, and 10.1 of Discrete Mathematics and Its Applications.

Introduction
Petri nets are mathematical structures which are useful in studying systems
such as computer hardware and software systems. A system is modeled by a
Petri net, the analysis of which may then reveal information about the structure
and behavior of the system as it changes over time. Petri nets have been used to
model computer hardware, software, operating systems, communications pro-
tocols, networks, concurrency, and parallelism, for example. Petri net models
have been constructed for the CDC6400 computer and the SCOPE 3.2 oper-
ating system for the purpose of evaluating system performance (see references
in [2]).
More generally, industrial production systems and even general social, eco-
logical, or environmental systems can be modeled by Petri nets. In this chapter
we will introduce the basic definitions of Petri net theory, study several exam-
ples in detail, and investigate some of the deeper concepts of the theory.

431
432 Applications of Discrete Mathematics

Petri Nets
A Petri net has four components: a set P of places, a set T of transitions,
an input function I, and an output function O. The input function assigns to
each transition a set of places known as the input places of the transition. The
output function assigns to each transition a set of places known as the output
places of the transition. Conceptually, places are passive components of the
system. They may store things, represent states of the system, or make things
observable, for example. Transitions are the active components of the system.
They may produce things, transport things, or represent changes of state of the
system. The following is the formal definition.

Definition 1 A Petri net structure is a four-tuple (P, T, I, O) such that P is

a finite set (of places); T is a finite set (of transitions), with P ∩ T = ∅; I (the
input function) is a function from T to the set of all finite subsets of P ; O (the
output function) is a function from T to the set of all finite subsets of P .
A place p is an input place for a transition t if and only if p belongs to I(t)
and p is an output place for t if and only if p belongs to O(t).

Example 1 Consider a simple batch processing computer system. We sup-

pose it has four states, or conditions:
1) a job is waiting for processing,
2) the processor is idle,
3) a job is being processed,
4) a job is waiting to be output.
There are four actions of the system:
1) a job is entered in the input queue,
2) a job is started on the processor,
3) a job is completed,
4) a job is output.
If a job is entered in the input queue, then a job is waiting; if a job is waiting
and the processor is idle, then a job can be started. After a job is processed it
can be completed and a job will be waiting for output and the processor will be
idle. If a job is waiting for output then a job can be output. Using the rule of
thumb about active and passive components of a system we should have places
corresponding to the four states of the system and transitions corresponding to
the four actions.
This simple system can be modeled by a Petri net having four places,
p1 , p2 , p3 , p4 , corresponding to the four states, and four transitions, t1 , t2 , t3 , t4 ,
corresponding to the four actions.
 Chapter 24 Petri Nets 433

The input and output functions I and O are defined by listing their values
explicitly:

I(t1 ) = ∅, I(t2 ) = {p1 , p2 }, I(t3 ) = {p3 }, I(t4 ) = {p4 }

O(t1 ) = {p1 }, O(t2 ) = {p3 }, O(t3 ) = {p2 , p4 }, O(t4 ) = ∅.

Note that I assigns t1 the subset ∅ of P and O assigns ∅ to t4 . This is the

formal way of saying that a transition has no input or output places.

There is a representation of Petri net structures using directed graphs,

which makes it easier to see the structure. In drawing the directed graph
representation of a Petri net, we make the following convention: places will be
represented by circular vertices and transitions by rectangular vertices. If a
place p is an input place for the transition t, then there is a directed edge from
the vertex for p to that for t. If p is an output place for t, then a directed edge
goes from t to p. It is easy to go back and forth between the two specifications.

Example 2 Construct a directed graph representation of the Petri net struc-

ture of Example 1.
Solution: Figure 1 shows such a directed graph and is labeled to show it as
a model for the simple batch processing computer system.

Figure 1. A Petri net graph for a computer system.

434 Applications of Discrete Mathematics

Example 3 As another example (adapted from [2]), consider a library sys-

tem. Users have access to three desks: the request desk, the collection desk,
and the return desk. All books are kept in the stacks and every book has an
index card. A user goes to the request desk to ask for a book. If the book is
in the stacks it is removed and the borrowed book index is updated. The user
gets the book at the collection desk. Books are returned to the return desk.
Books returned are replaced in the stacks and the index is updated. A Petri
net model of this system is given in graphical form in Figure 2.

Figure 2. A library system.

For the purpose of modeling more complex systems it is necessary to use

a more complicated graph structure in which more than one directed edge may
go from one vertex to another.

Example 4 Figure 3 shows an example of a graphical Petri net with multiple

edges between some vertices.

Figure 3. A graphical Petri net.

This sort of structure requires a modification of the definition of Petri net

structure given previously since a place cannot appear more than once as a
member of a set of input or output places. To overcome this difficulty we need
only change the ranges of the input function I and the output function O to
be the set of all finite subsets of the Cartesian product of P and N , the set of
 Chapter 24 Petri Nets 435

positive integers. Thus an input or output place would be represented as an

ordered pair (p, n) where the place is p and n is the number of edges between p
and the transition in question. In what follows we will generally be using only
the graphical representation of the net.

Modeling System Dynamics

In addition to having a static structure which is modeled by Petri net structures
as discussed above, many systems change over time and it is of great interest to
study this dynamic behavior. In Petri net theory this is accomplished through
the use of markings.

Definition 2 A marking of a Petri net (P, T, I, O) is a function m from P

to the nonnegative integers.

We think of the marking as assigning tokens to the places in the net. The
number assigned to a place by the marking is the number of tokens at that
place. In models of systems the tokens could have many interpretations. For
example in modeling computer operating systems the tokens might represent
the processes which compete for resources and which must be controlled by the
operating system.
As we shall see, the tokens can move from place to place in the net simu-
lating the dynamic behavior of the system. In the graphical representation of
Petri nets the tokens are represented by solid dots in the circular vertices for
the places. This is practical when the number of tokens at any place is small.
If there are too many tokens then we write the number of tokens assigned to a
place inside the circle.

Example 5 Figure 4 shows some examples of marked Petri nets.

Figure 4. Marked Petri nets.

436 Applications of Discrete Mathematics

The dynamic change of the Petri net model is controlled by the marking,
i.e., the number of tokens assigned to each place, and by the firing rules for
transitions.

Definition 3 A transition is enabled for firing if each of its input places has
at least as many tokens in it as edges from the input place to the transition. A
transition may fire only if it is enabled.

Example 6 Which of the transitions in Figure 5 are enabled for firing?

Figure 5. Transitions.

Solution: Transitions a and b are enabled while c and d are not.

A transition is fired by removing one token from each input place for every
edge from the input place to the transition and adding one token to each output
place for every edge from the transition to that place.

Example 7 Show the result of firing each of the transitions in Figure 6.

Figure 6. Transitions before firing.

Solution: Figure 7 shows these transitions after firing.
 Chapter 24 Petri Nets 437

Figure 7. Transitions after firing.

Firing a transition generally changes the marking of the Petri net to a

different one. Thus the dynamics of the system are acted out on the markings.
Note that since only enabled transitions may be fired, the number of tokens
at each place is always non-negative. Note also that firing transitions does not
necessarily conserve the total number of tokens in the net.

We may continue firing transitions until there are no more enabled for
firing. Since many transitions might at any time be enabled for firing, the
sequence in which the transitions are fired might not be unique. The only
restriction is that transitions do not fire simultaneously. Consequently, the
marking which results after firing a certain number of transitions could vary
depending on the particular transitions fired and their order.

Thus, there is possibly some nondeterminism in the execution of Petri net

firing. Figure 8 shows an example of two transitions, each of which is enabled
for firing, but once one of them is fired, the other is no longer enabled.

Figure 8. Transitions enabled for firing.

The goal is to have the Petri net model the system, so each possible firing
sequence would correspond to a possible behavior of the system, even though
they would not all necessarily result in the same final marking.
438 Applications of Discrete Mathematics

Applications
We now look at some examples of Petri net models. The models we consider
in detail all concern problems related to computer hardware and operating
systems.

Example 8 Construct a Petri net model for calculating the logical conjunc-
tion of two variables x and y, each of which takes the values “true” or “false”.
Each is assigned a value independently of the other.
Solution: Figure 9 shows the required net. To see how this net works, we
walk through the firing sequence corresponding to x = “true” and y = “false”.
We start with one token in the places corresponding to x =“true” and y =
“false”, the leftmost and rightmost places on the top line of Figure 9. Now only
one transition is enabled for firing, the second one from the left. Firing that
transition puts one token into the place labeled x ∧ y = “false”. No further
transition firings are possible and the desired result is obtained. The reader
may check that other choices for values of x and y also yield the correct results.

Figure 9. Petri net model for x ∧ y.

Example 9 Construct a Petri net that computes the negation of x, which

again takes only the values “true” and “false”.
Solution: Figure 10 shows a net that computes the negation of the variable x.
This net works similarly to that in Figure 9. If we want to know the value of
¬x when x has a certain value, we put one token in the place for that value,
carry out the firing, and observe which place has a token in it.
 Chapter 24 Petri Nets 439

Figure 10. Petri net model for ¬x.

Petri nets can be combined to form new nets. One way is simply to take
the disjoint union of the two nets. This is accomplished for the graphical rep-
resentation by simply drawing the graphs next to each other. Another method
combines the nets in a way similar to the composition of functions. In many
nets certain of the places could naturally be considered as input places, for
example those labeled x = “true” and x = “false” in Figure 10. Others could
be considered as output places, for example those labeled ¬x = “false” and ¬x
= “true” in Figure 10. To compose two nets N1 and N2 , the number of output
places of N1 must equal number of input places of N2 . The composition is
obtained by identifying the output places of N1 with the input places of N2 in
an appropriate manner.

Example 10 Construct a Petri net for computing ¬(x ∧ y) by composing

the nets in Figures 9 and 10.
Solution: Figure 11 shows the composition net. The input places for Figure 9
are the places labeled x = “true”, x = “false”, y = “true”, and y = “false”. The
output places in Figure 9 are the places labeled x ∧ y = “true”, x ∧ y = “false”.
The input places for Figure 10 are those labeled x = “true” and x = “false”.
To combine two nets we match up the input and output places as shown. It
is also necessary to relabel some of the places to indicate their meaning in the
composed net.

Now that we have seen how to use Petri nets to model conjunction and
negation of propositional formulas we should note that we can represent all
propositional functions using Petri nets since conjunction and negation are func-
tionally complete, i.e. every possible truth table can be obtained as that of a
propositional function using only conjunction and negation. (See Section 1.2 of
440 Applications of Discrete Mathematics

Figure 11. Composition net for nets in Figures 9 and 10.

Discrete Mathematics and Its Applications.)

Using the correspondence between propositional functions and switching
circuits we can observe that computer hardware can also be modeled using Petri
nets. Figure 11 can thus be seen as modeling a NAND gate.
Our next examples of Petri nets show how they can be useful in studying
parallelism in computer systems or problems of control in operating systems.
First we will examine the readers/writers problem. There are two types of
processes: reader processes and writer processes. All processes share a common
file or data object. Reader processes never modify the object but writer pro-
cesses do modify it. Therefore, writer processes must exclude all other processes,
but multiple reader processes can access the shared object simultaneously. This
sort of thing happens frequently in database applications, for example, in air-
line reservation systems. Depending on the actual system, there may be limits
on how many processes may access the memory area simultaneously. How can
this situation be controlled?

Example 11 Construct a Petri net model for the readers/writers problem

 Chapter 24 Petri Nets 441

when there are six processes, two of which have write access and four have read
access. Furthermore, suppose at most three reader processes can overlap in
access to memory.
Solution: Figure 12 shows a Petri net model for this control problem. This
example was adapted from one in reference [2]. In this example the processes
are represented by the tokens, the places in the net represent the different things
the processes might be doing, or states they might be in.

Figure 12. Petri net model for readers/writers problem.

The assignment of tokens to places in Figure 12 is just one possibility, for
illustrative purposes. Let us follow this system through some transition firings
to see how it works. As the system is in Figure 12, no processes are reading
or writing. Transition (a) is enabled for firing, as are (d), (e), and (b). If we
fire (a) then one token is put into the writing place, the place labeled “ready to
write” has one token removed from it, and the “access control” place has three
tokens removed from it. After this firing, transition (d) is no longer enabled,
so no process can enter the “reading” place. Even if there were another token
in the “ready to write” place, we could not fire (a) again because there are no
tokens in the access control place so (a) is not enabled. In fact, a little thought
shows that if (a) has been fired to put a process in the write place then (a)
and (d) become unable to fire, thus preventing any other process from either
reading or writing. Now transition (c) is enabled. If it is fired we will have two
tokens in the “other processing” place and three tokens will be put back in the
access control place.
We now see how the net controls read access. Transition (e) is enabled for
firing, and in fact we can fire it three times in succession so that the reader
“other processing” place becomes empty and the “ready to read” place has
four tokens in it. Transition (d) is now enabled for firing. Each time we fire
transition (d), one token is removed from the “ready to read” place, one token
is put in the reading place, and one token is taken out of the access control
place. Once three firings of (d) have taken place, then the access control place
442 Applications of Discrete Mathematics

is empty, transition (d) is no longer enabled, and no additional tokens can be

put in the reading place.
What happens when processes stop reading? Every time transition (f) is
fired, one token is put in the access control place and one token is put in the
“other processing” place. Three firings of (f) will put three tokens into access
control and if there are any tokens in the “ready to write” place, transition (a)
will be enabled. We can see how the three tokens in the access control place
implement the requirement that no more than three reader processes may have
access simultaneously, and the three edges between transitions (a) and (e) and
the access control place prevent access when a process is writing.

Next we examine the producer/consumer problem. Again we have a shared

data object, but this time it is specified to be a buffer. A producer process
creates objects which are put in the buffer. The consumer waits until there is
an object in the buffer, removes it, and consumes it.

Example 12 Construct a Petri net model for the Producer/Consumer prob-

lem.
Solution: The net in Figure 13 is the required model.

Figure 13. Petri net for producer/consumer problem.

In this model each token in the buffer represents an object which has been
produced but not yet consumed. We can modify this example to work for the
multiple producer/multiple consumer problem. In this case multiple producers
 Chapter 24 Petri Nets 443

produce objects which are stored in a common buffer for the use of multiple
consumers. If there are n producers and k consumers, we need only start the
system with n tokens in the initial place of the producer process and k tokens
in the initial place of the consumer process, instead of one in each as shown in
Figure 13.

The dining philosophers problem was suggested in 1968 by Edsger Dijkstra,

a great pioneer in computer science. Five philosophers alternatively think and
eat. They are seated around a large round table on which are a variety of
Chinese foods. Between each pair of philosophers is one chop stick. To eat
Chinese food, two chopsticks are necessary; hence each philosopher must pick
up both the chopstick on the left and that on the right. Of course, if all the
philosophers pick up the chopstick on their right and then wait for the chopstick
on their left to become free, they will wait forever — a deadlock condition.
Dijkstra formulated this problem to illustrate control problems that confront
operating systems in which processes share resources and may compete for them
with deadlock a conceivable result. To solve the problem some philosophers
must eat while others are meditating. How can this be accomplished?

Example 13 Construct a Petri net to solve the problem of the dining

philosophers.
Solution: The Petri net in Figure 14 solves this problem.
Each philosopher is represented by two places, meditating (Mi ) and eating
(Ei ). There are also two transitions for each philosopher for going from medi-
tating to eating and vice-versa. Each chopstick is represented by a place (Ci ).
A token in an eating or meditating place indicates which condition that philoso-
pher is in. A token in a chopstick place indicates that that chopstick is free. A
philosopher can change from meditating to eating only if the chopsticks on the
left and right are both free.

Numerous other applications of Petri net models have been developed, not
all in computer science. Petri nets have been used to study the PERT technique
of project planning and in studying legal processes. More information on these
topics can be found in reference [1]. The bibliography of reference [2] lists 221
items including 47 papers and books on applications of Petri nets.

Analysis of Petri Nets

We have seen how useful Petri nets can be in constructing models, but perhaps
444 Applications of Discrete Mathematics

Figure 14. Petri net for dining philosophers problem.

the most important purpose of modeling is to be able to analyze the model

to answer questions about the system being modeled. In this section we con-
sider some problems related to the analysis of Petri nets. The most important
analysis problem is that of reachability.

Definition 4 Given a Petri net and a marking m, a marking m is immedi-

ately reachable from m if there is some transition which, when fired, yields the
marking m .
A marking m is reachable from the marking m if there is a sequence
m1 , . . . , mk of markings such that m1 = m and mk = m and for all i, mi+1 is
immediately reachable from mi . If N is a Petri net and m is a marking of N ,
then the set of all markings reachable from m is called the reachability set of
m and is denoted R(N, m).

Example 14 Show that the marking of the net in Figure 15b is reachable
from the marking in Figure 15a.
Solution: Firing t1 twice and then t2 yields the marking in Figure 15b.
 Chapter 24 Petri Nets 445

Figure 15. A reachable net.

Many interesting analysis problems for Petri nets can be phrased as ques-
tions about what kinds of markings are reachable from others. For example we
introduced the concept of deadlock in the discussion of the dining philosophers
problem. A net is in deadlock if no transitions can fire. In any net we might
be interested in knowing whether a deadlock marking is reachable from some
initial marking.
If a Petri net is to be a model of a real hardware device, one of the important
properties it should have is safeness.

Definition 5 A place p in a Petri net N with marking m is safe if and only

if for all m ∈ R(N, m), m (p) ≤ 1. That is, p never has more than one token.
A net is safe if every place in it is safe.

The reason safeness is important in modeling hardware devices is that if a

place has either no tokens or one token in it, then that place can be implemented
by a flip-flop. The nets in Figures 9, 10, and 11 are safe if the initial marking
puts at most one token in each of the initial places. The producer/consumer
net of Figure 13 is also safe with the marking shown.
Safeness is a special case of the property called boundedness. A place is
k-bounded if the number of tokens in that place can never exceed k. A net
is k-bounded if all of its places are k-bounded. Safeness is thus the same as
1-boundedness. A Petri net is bounded if it is k-bounded for some k. A Petri
net which is bounded could be realized in hardware using counters for places
while one with an unbounded place could not.
Petri nets can be used to model resource allocation systems. In this context
the tokens might represent the resources to be allocated among the places. For
these nets conservation is an important property.

Definition 6 A Petri net N with marking m is conservative if for all m ∈

446 Applications of Discrete Mathematics

R(N, m),

 
m (p) = m(p).
p∈P p∈P

Conservation thus means that the total number of tokens is the same for
all markings reachable from m. This is an extremely strong requirement, as the
following theorem shows.

Theorem 1 If N is a conservative net then for each transition the number

of input places must equal the number of output places.
Proof: If t is a transition with differing numbers of input and output places
then firing t will change the number of tokens in the net. (See, for example,
the second transition in Figure 6.)

The concepts presented so far all involve reachability in some sense. How
can they be analyzed? Numerous techniques have been developed and are
discussed in detail in references [1] and [2]. We will limit ourselves here to a
discussion of the reachability tree. The reachability tree of a Petri net is a
graphical method of listing the reachable markings. Our treatment follows that
of reference [1], Chapter 4, Section 2. To begin, we fix an ordered listing of the
places so that a marking can be written as an ordered n-tuple of nonnegative
integers. An entry in an n-tuple gives the number of tokens assigned to the
place in that position in the list. The initial marking of the net corresponds
to the root vertex of the tree. For each marking immediately reachable from
the initial one, we create a new vertex and draw a directed edge to it from the
initial marking. The edge is labeled with the transition fired to yield the new
marking.
Now we repeat this process for all the new markings. If this process is
repeated over and over, potentially endlessly, all markings reachable from the
initial one will eventually be produced. Of course, the resulting tree may well
be infinite. Indeed, if the reachability set of a net is infinite, then the tree must
also be infinite.

Example 15 Carry out the first three steps in construction of the reacha-
bility tree for the net shown in Figure 16, with the initial marking given there.
Solution: Figure 17 shows the result. The tree is obtained as follows: the
triple (x1 , x2 , x3 ) gives the number of tokens assigned to the places p1 , p2 , p3 ,
in that order. From marking (1, 0, 0), firing t1 yields marking (1, 1, 0). From
marking (1, 0, 0) firing t2 gives marking (0, 1, 1). From marking (1, 1, 0), firing
 Chapter 24 Petri Nets 447

Figure 16. A Petri net.

t1 yields (1, 2, 0) while firing t2 yields (0, 2, 1). From (0, 1, 1), firing t3 yields
(0, 0, 1). This completes the second level down from the top. The third level is
constructed similarly.

Figure 17. First three steps of reachability tree for Petri net
of Figure 16.
If a net has an infinite reachability set, then of course its reachability tree
will be infinite. However, it could happen that a net has a finite reachability set
and still has an infinite reachability tree. This would be the case if a sequence
of markings is repeated infinitely as in the net in Figure 18. For this net the
tree is an infinite chain since the net alternates between the markings (1, 0)
and (0, 1). As we have constructed it, the reachability tree contains the results
of every possible sequence of transition firings, each sequence corresponding to
some path from the root. Any reachability problem can, then, be solved by
examining the reachability tree, though the search may be impractical.
While the reachability tree contains every reachable marking, its potential
infiniteness poses a problem for analysis. Consequently, it is useful to have a
finite representation of the reachability tree. We now describe a method for
obtaining a finite representation of a reachability tree. For this purpose we
introduce a classification of vertices in the tree. The new markings produced
448 Applications of Discrete Mathematics

Figure 18. Petri net with infinite reachability tree.

at each step will be called frontier vertices. Markings in which no transition

is enabled will be called terminal vertices. Markings which have previously
appeared in the tree will be called duplicate vertices.
Terminal vertices are important because they yield no additional frontier
vertices. Duplicate vertices yield no new frontier vertices, so no successors of
duplicate vertices need be considered. We can thus stop generating frontier
vertices from duplicate ones, thereby reducing the size of the tree. For the
example in Figure 18 this will make the tree finite. In the example of Figures 16
and 17, the vertex (0, 1, 1) in the third level from the root is a duplicate vertex
and hence will produce no new markings.
One more case needs to be considered. If we examine the example of
Figures 16 and 17 we can see that transition t1 can be fired over and over
endlessly, each time increasing the number of tokens at place 2 by one. This
will create an infinite number of different markings. In this situation and in
others like it there is a pattern to the infinite string of markings obtained. To
aid in describing this pattern, it is useful to regard the n-tuples for markings
as vectors and perform operations on them component-wise.
Suppose we have an initial marking m and some sequence of transition
firings s yields a marking m with m ≥ m componentwise. We can think of m
as having been obtained from m by adding some extra tokens to some places
so
m = m + (m − m)
componentwise, and m − m ≥ 0. Thus, firing the sequence s had the result of
adding m − m to m. If we start from m and fire s to get m , then the result
will be to add m − m to m , so

m = m + 2(m − m).

In general we could fire the sequence n times to get a marking m + n(m − m).
 Chapter 24 Petri Nets 449

We use a special symbol, ω, to represent the infinite number of markings

which result from this type of loop. It stands for a number of tokens which can
be made arbitrarily large. For the sake of simplifying our description it is useful
to define some arithmetic “operations” using ω. For any constant a we define

ω+a=ω ω−a=ω a<ω ω ≤ ω.

These are all we will need in dealing with the reachability tree. We need one
more notational device to simplify the description of the method for construct-
ing the reduced reachability tree.

Definition 7 If m is a marking of a Petri net and t is a transition enabled

for firing, then δ(m, t) is the marking produced from m by firing t.

In the reduced reachability tree each vertex x is an extended marking; that

is, the number of tokens at a place is allowed to be either a nonnegative integer
or the symbol ω. Each vertex is classified as either a frontier vertex, a duplicate
vertex, a terminal vertex, or an interior vertex. Frontier vertices are vertices
which have not been processed by the reduction algorithm; they are converted
to terminal, duplicate, or interior vertices.
The algorithm begins by defining the initial marking to be the root of
the tree and initially a frontier vertex. The algorithm continues until no more
frontier vertices remain to be processed. If x is a marking written as an n-tuple,
let xi be the ith component of x.
Let x be a frontier vertex to be processed.
1. If there exists another vertex in the tree which is not a frontier vertex
and is the same marking as x, then vertex x is a duplicate vertex.
2. If no transitions are enabled for the marking x, then x is a terminal
vertex.
3. For all transitions tj which are enabled for firing for the marking x,
create a new vertex z in the tree. The components of z are defined as follows:
(a) If xi = ω then zi = ω.
(b) If there exists a vertex y on the path from the root to x with
y ≤ δ(x, tj ) and yi < δ(x, tj )i then zi = ω.
(c) Otherwise, zi = δ(x, tj )i . An edge labeled tj is directed from vertex
x to vertex z. Vertex x is classified as an interior vertex and z becomes a frontier
vertex.
When there are no more frontier vertices, the algorithm stops.

Example 16 Construct the reduced reachability tree for the Petri net shown
in Figure 16.
450 Applications of Discrete Mathematics

Solution: Figure 19 shows the required tree.

Figure 19. Reduced reachability tree.

It has been proved that the algorithm for constructing the reduced reach-
ability tree always terminates in a finite tree. The interested reader is referred
to Chapter 4 of [1] for details.
Many questions about reachability can be answered using the reduced
reachability tree. However, because the use of the symbol ω condenses infinitely
many vertices into one, some information may be lost by this process. Often
the ω indicates a pattern which can be recognized, thus reducing the loss of
information. The special reachability questions we presented can be answered
using the tree. A net will be safe if the vertices in the tree all have only 0s
and 1s as components. A net will be bounded provided the symbol ω never
appears. A net will be conservative if the sums of the components of all the
vertices are equal.
Petri nets are perhaps the most widely used modeling tool in computer
science. The references each contain large bibliographies; [1] has a bibliography
extending over 38 pages.

Suggested Readings

1. J. Peterson, Petri Net Theory and the Modeling of Systems, Prentice Hall,
Upper Saddle River, N.J., 1981.

2. W. Reisig, Petri Nets, An Introduction, Springer-Verlag, New York, 1985.

 Chapter 24 Petri Nets 451

Exercises
1. Construct a graph representation for the Petri net with places {p1 , p2 , p3 , p4 ,
p5 }, transitions {t1 , t2 , t3 }, and input function I and output function O
defined by: I(t1 ) = {p1 }, I(t2 ) = {p1 }, I(t3 ) = {p2 , p3 }, O(t1 ) = {p2 },
O(t2 ) = {p3 }, O(t3 ) = {p4 , p5 }.
2. A small machine shop can make three different items, though it can work
on only one item at a time. The shop can be in six different “states”
corresponding to: an order is waiting, one of the three items is being made,
an order is finished, and the shop is idle. There are six actions: order
arriving, start work on item 1, start work on item 2, start work on item 3,
finish processing, and order sent for delivery. Construct a Petri net model
for this machine shop analogous to that in Figure 1.
3. Determine whether each transition is enabled for firing.
a) b)

c)

4. Find the results of firing each transition.

a) b)

c)
452 Applications of Discrete Mathematics

5. Construct a Petri net to calculate the propositional function x ∨ y.

6. Construct a Petri net to calculate the propositional function ¬x ∧ ¬y.

7. Construct a Petri net for the propositional function x → y.

8. Modify the example of Figure 12 to allow for three writer processes, five
reader processes, and an access limit of only two reader processes at a time.

9. Construct a Petri net for the producer/consumer problem in which there

are two producers and three consumers.

10. In our solution of the producer/consumer problem there is no limit on the

number of tokens that could be in the buffer. Show how to modify the net
so that at most three tokens can be in the buffer.

11. Consider the following marked Petri net.

a) Is the net safe?
b) Is the net bounded?
c) Is the net conservative?

12. Analyze the following Petri net for safeness, boundedness, and conserva-
tiveness.

13. Construct the reachability tree for the following Petri net.
 Chapter 24 Petri Nets 453

a) Is the net safe?

b) Is the net conservative?
c) Is the marking (0, 0, 1, 0, 1) reachable from the initial marking?
14. Construct the first four levels of the reachability tree of the following Petri
net. Is this net bounded?

15. Construct the reduced reachability tree for the Petri net in Exercise 14.

Computer Projects

1. Implement computer models of Petri nets. That is, write a program which
would simulate the behavior of a marked Petri net under transition firing.
2. Write a computer program to calculate the reachability tree up to ten levels
(if necessary) for a Petri net.
3. Write a computer program to calculate the reduced reachability tree of a
Petri net with a given marking.