The Theory of Automata In this chapter we begin with the study of automaton. We deal with transition systems which are more general than finite automata. We define the acceptability of strings by finite automata and prove that nondeterministc finite automata have the same capability as the deterministic automata as far as acceptability is concemed. Besides, we discuss the equivalence of Mealy and Moore models, Finally, in the last section. we give an algorithm to construct a minimum state automaton equivalent 10 4 given finite automaton, 3.1 DEFINITION OF AN AUTOMATON We shall give the most general definition of an automaton and later modify it to computer applications. An automaton is defined as a system where energy, materials and information are transformed, transmitted and used for performing some functions without direct participation of man. Examples are ‘automatic machine tools, automatic packing machines. and automatic photo printing machines, In computer science the term ‘automaton’ means ‘discrete automaton’ and is defined in a more abstract way as shown in Fig. 3.1 4 Automaton |-——+ 0, au teins Log Fig. 3.1 Model of a discrete automaton, 7The characteristics of automaton are now described. (i) Input. At each of the discrete instants of time 1). fy... fy the input values 1h. Is, «++ yp each of which can take a finite number of fixed values from the input alphabet £, are applied to the input side of the model shown in Fig. 3.1 (ii) Ouput. 04, Os, .... Oy are the outputs of the model, each of which can take a finite number of fixed values from an output 0. (ii) States. At any instant of time the automaton can be in one of the SALES iy Gy Me Gy) State relation. The next state of an automaton at any instant of time is determined by the present state and the present input, (\) Output relation. The output is related to either state only or to both the input and the state. It should be noted that at any instant of time the automaton is in some state. On ‘reading’ an input symbol, the automaton moves to a next state which is given by the state relation Note: An automaton in which the output depends only on the input is called {an automaton without a memory. An automaton in which the ousput depends ‘onthe states as well is called automaton with a finite memory. An automaton in which the output depends only on the states of the machine is called a Moore machine. An automaton in which the output depends on the state as well as on the input at any instant of time is called a Mealy machine. Consider the simple shift register shown in Fig, 3.2 as a finite-state machine and study its operation. Serial input > pe Ps > Fig. 3.2 A 4-bit serial shift register using D flip-fops. Solution ‘The shift register (Fig. 3.2) can have 2* = 16 states (0000, 0001, .... LID). and one serial input and one serial output. The input alphabet is Z = (0. 1). and the output alphabet is O = (0. 1}. This 4-bit serial shift register can be further represented as in Fig. 3.3.Chapter 3: The Theory of Automata Si 73 } ange Fig. 3.3. A shift register as 2 finite-state machine From the operation. it is clear that the output will depend upon both the input and the state and so it is a Mealy machine. Tn general. any sequential machine behaviour can be represented by an automaton, 3.2 DESCRIPTION OF A FINITE AUTOMATON Definition 3.1 Analytically. a finite automaton can be represented by a Setuple (Q. E. 6. go. FY. where () Q isa finite nonempty: set of states Gi) E is a finite nonempty set of inputs called the inpur alphaber. (iii) dis a function which maps Q into Q and is usually ealled the direct transition fiction. This is the function which describes the change of states during the transition. This mapping is usually represented by a transition table or a transition diagram. (iv) qo € Q is the initial state. () F< Qis the set of final states. It is asstumed here that there may be more than one final state Note: The transition function which maps Q x E* into Q (ie. maps a state and a string of input symbols including the empty string into a state) is ealled the indirect sransition function. We shall use the same symbol 6 to represent both types of transition functions and the difference can be easily identified by the nature of mapping (symbol or a string). ie. by the argument. 6 is also called the nest state function. The above model can be represented graphically by Fig. 34. iaeeeeeeeed/tiescea: f ] TTT | | c ] | input | LL | tape —— Reading head Fig. 3.4 Block diagram of @ finite automaton74 Si Theory of Computer Science Figure 3.4 is the block diagram for a finite automaton, The various components are explained as follows: (i) Input tape. The input tape is divided into squares. each square containing a single symbol from the input alphabet E. The end squares Of the tape contain the endmarker q at the left end and the end- marker § at the right end. The absence of endmarkers indicates thai the tape is of infinite length. The left-to-right sequence of symbols between the two endmarkers is the input string to be processed. (ii) Reading head. ‘The head examines only one square at a time and can ‘move one square either to the left or to the right. For further analysis ‘we restrict the movement of the R-head only to the right side ii) Finite control. The input to the finite control will usually be the symbol under the R-head, say a, and the present state of the machine, say g, to give the following outputs: (a) A motion of R-head along the tape to the next square (in some a null move, ie. the R-head remaining to the same square is permitted); (b) the next state of the finite state machine given by 5(q, a). 3.3. TRANSITION SYSTEMS Actransition graph or a transition system is a finite directed labelled graph in which each vertex (or node) represents a state and the directed edges indicate the transition of a state and the edges are labelled with imput/output. A typical transition system is shown in Fig. 3.5. In the figure, the initial state is represented by a circle with an arrow pointing towards it, the final state by two concentric circles, and the other states are represented by just a circle. ‘The edges are labelled by input/output (e.g. by 1/0 or 1/1). For example. ifthe system is in the state gy and the input 1 is applied, the system moves to state 4) as there is a directed edge from gy to gy with label 1/0, It outputs 0. oo 10 wn Qn % 00 Fig. 35 A transition system ‘We now give the (analytical) definition of a transition system, Definition 3.2 A transition system is a S-tuple (Q, E, & Qp, F), where (i) Q, ¥ and F are the finite nonempty set of states, the input alphabet, and the set of final states, respectively, as in the case of finite automata; (ii) Qy © Q and Qp is nonempty: and ii) Sis a finite subset of Q x E° x Q.Chapter 3: The Theory of Automata 8. 75 In other words, if (gy, q3) is in 6. it means that the graph starts at the vertex qy, goes slong a set of edges, and reaches the vertex qs. The concatenation of the label of all the edges thus encountered is ws: Definition 3.3 A transition system accepts a string w in E* if (i) there exists @ path which originates from some initial state, goes along the arrows, and terminates at some final state; and Gi) the path value obtained by concatenation of all edge-labels of the path is equal t0 w Eee Consider the transition system given in Fig. 3.6. 0 Fig. 3.6 Transition system for Example 3.2. Determine the initial states. the final states. and the acceptability of 101011, 111010. Solution The initial states are qo and qy. There is only one final state, namely ¢ The path-value of goqegods is 1OI0LL, AS qs is the final state, 101011 is accepted by the transition system, But, 111010 is not accepted by the transition system as there is no path with path value 111010. : Every finite automaton (Q, qo. F) can be viewed as a transition system (Q, £, 6” Qp, F) if we take Qy = {go} and 8’ = [(g, w, dg, wg € Q, w @ E*}, But a transition system need not be a finite automaton. For example. a transition system may contain more than one initial state. 3.4 PROPERTIES OF TRANSITION FUNCTIONS Property 1 6(q, A) = q is finite automaton, This means that the state of the systemn can be changed only by an input symbol.76 Theory of Computer Science Property 2 For all strings w and input symbols « 6iq, an) = 6, a), W) 6(q, wa) = 65g, w), a This property gives the state after the automaton consumes or reads the first symbol of @ siring aw and the state after the automaton consumes a prefix of the string wa, Sze Prove that for any transition function & and for any two input strings x and y; 6g, wy) = 6(8(q. oY) GA) Proof By the method of induction on |y’|, ie. length of »: Basis: When |y|= Ly sae = LHS. of B.)= 5g. x0) = 5(6(q, 9, a) by Property 2 RHS. of 3.1) Assume the result, ic. (3.1) for all strings x and strings y with [| = a. Let y be a string of length m+ 1. Write y= yy@ where [yy| = 1 LHS. of (G1) = 6g, sya) = 8g. x, ey 68g, 8). a) by Property 2 45g, 0). Sq, 2), 1), a) by induction hypothesis RES. of G1) = 65q. 01 9) = GHG, yy), aby Property 2 Hence, LHS. = RHS. This proves (3.1) for any string y of length m + 1. By the principle of induction, (3.1) is true for all strings, 1 EXAMPLE 3.4 Prove that if 6(q, x) = 6(g, »), then 5¢g, 6(q, 32) for all strings < in E* Solution 6a, (6G, ». by Example 3.3 851g, », © 82 By Example 3.3. iq, 5) = HBLG, J. 2 = 8g, x Ba)Chapter 3: 3.5 ACCEPTABILITY OF A STRING BY A FINITE AUTOMATON Definition 3.4 A string « is accepted by a finite automaton M=(Q.E. 8 40 FY if igo. x) = @ for some qe F. This is basically the acceptability of a string by the final state Note: A final state is also called an accepting state EXAMPLE 3. Consider the finite st i¢ machine whose transition function is given by Table 3.1 in the form of a transition table. Here. Q = {go 41, 2 dsl. E = (0, 1) F = (qo). Give the entire sequence of states for the input string 110001 TABLE 3.1. Transition Function Table for Example 3.5 State 5 5 7 & 3 4 ® a @ % @ 4 s & Solution L 4 &qy. HOLO1) = dq. 10101) = 6(qo,0101) 4 = digs 101) 1 = 5iq3.01) L lq) = 5g A) =a Hence. La kg & ae 0 > > Io 2 > 49 > dy The symbol 4 indicates that the current input symbol is being processed by the machine78 ©. Theoryof Computer Science 3.6 NONDETERMINISTIC FINITE STATE MACHINES We explain the concept of nondeterministic finite automaton using a transition diagram (Fig. 3.7) 3.7. Transition system representing nondeterministic automaton. Tf the automaton is in a state {qo} and the input symbol is 0, what will be the next state? From the figure itis clear that the next state will be either (4o} or {qu}. Thus some moves of the machine cannot be determined uniquely by the input symbol and the present state, Such machines are called nondeterministie automata. the formal definition of which is now given Definition 3.5 A nondeterministic finite automaton (NDFA) is a S-tuple (3.6 do, F), where (] Qis a finite nonempty set of states: (i) Z is a finite nonempty set of inputs Gi) 3s the transition function mapping from Q x & into 2 which is the power set of @, the set of all subsets of Q: iv) go © Q is the initial stare: and (8) FS Qs the set of final states. We note that the difference between the deterministic and nondeterministie automata is only in 6. For deterministic automaton (DEA), the outcome is a state, ie, an element of Q; for nondeterministic automaton the outcome is a subset of @. ‘Consider, for example, the nondeterministic automaton whose transition diagram is described by Fig. 3.8. ‘The sequence Uf sates for die input suing 0100 is givew in Fig. 3.9. Hence, qu 0100) = {90 4x da} Since q, is an accepting state, the input string 0100 will be accepted by the nondeterministic automaton.Chapter 3: The Theory of Automata S79 %—* % ——+ a) —2 9 — % VN NY \ : Fig. 3.9 States reached while processing 0100. Definition 3.6 A string w = Z* is accepted by NDFA M if (qo, ) contains some final state Note: As M is nondeterministic. 5(gq, ) may have more than one state. So ww is accepted by M if a final state is one among the possible states that M can reach on application of We can visualize the working of an NDFA M as follows: Suppose M reaches a state q and reads an input symbol a. Tf 6g, a) has » elements the automaton splits inom identical copies of itself: each copy pursuing fone choice determined by an element of 5g, a). This type of parallel computation continues. When a copy encounters (g. a) for which 6(q, a) = 0, this copy of the machine ‘dies’; however the computation is pursued by the other copies. If any one of the copies of M reaches a final state after processing the entire input string w. then we say that M accepts w. Another ‘way of looking at the computation by an NDFA M is to assign a tree structure for computing 6(q, w). The root of the tree has the label g. For every input symbol in ve. the tree branches itself. When a leaf of the tree has a final state as rts label. then M accepts Definition 3.7 The set accepted by an automaton M (deterministic or nondeterministic) is the set of all input strings accepted by M. It is denoted by TM).80_& Theory of Computer Science 3.7 THE EQUIVALENCE OF DFA AND NDFA We naturally try to find the relation between DFA and NDFA. Intuitively we now feel that (i) A DPA can simulate the behaviour of NDFA by increasing the number of states. (In other words, a DFA (Q, 3, & qo F) can be viewed as an NDFA (Q, 3, 8", qo. F) by defining 5(q, a) = (51g, «)}.) (i) Any NDFA is a more general machine without being more powerful We now give a theorem on equivalence of DFA and NDFA. ‘Theorem 3.1 For every NDFA, there exists a DFA which simulates the ‘behaviour of NDFA. Alternatively, if J is the set accepted by NDFA, then there exists a DFA which also accepts L. Proof Let M=(Q. . & qo F) be an NDFA accepting L. We construct a DFA Mas: MOE 6 Ge PY where ( @ = 22 (any state in Q" is denored by [a qo. ++ Gh Where gy 4 4g = Qy Gi) a'y = Lgok: and (iil) F* is the set of all subsets of Q containing an element of F, Before defining 5% let us look at the construction of Q', gy and F’. M is initially at go, But on application of an input symbol, say a, M can reach any of the states 6(qu. a). To describe M, just after the application of the input symbol a, we require all the possible states that Mf can reach after the application of a. So, M”has to remember all these possible states at any instant of time. Hence the states of MM’ are defined as subsets of Q. As M starts with the initial state qo. 4% is defined as [qq]. A string w belongs to T(M) if a final siate is one of the possible states that M reaches on processing w: So, a final state in M’ (ie. an element of F’) is any subset of Q containing some final state of M. Now we can define 8% Gv) Big. 4 «gh. a) = 6(q, a) U Bly aU... U BGs a) Equivalently. Sli. gm vom gi = Be ve Bl if and only if Slay... 6 Gib a) = [Py Ps Pit Before proving L = T(M’), we prove an auxiliary result 8d = [d. «- + ah Gay if and only if 8iqo, x) = {q). --» qi} for all x in E* We prove hy induction on [x]. the “iP par, Sai = Lae geo Gd Bs) if ig. 3) = 4)Chapter 3: The Theory of Automata S81 When |x| = 0. S(qy A) = {ay}. and by definition of 6; d(qj, A) = 46 = [ao]. So. (3.5) is true for x with |x] = 0. Thus there is basis for induction ‘Assume that (3.5) is true for all strings y with |v] $m. Let x be a string of length m + 1, We can write x as va, where |y| =m and ae E Let (go. 9) = (Pie oe pi and Blqy. va) = (rr, vo Tu) As [yh Sm by induction hypothesis we have Sigh. YY = IPv Pil 3.6) Also. (ry Py ri) = lq, Ya) = HSlqa. 99. a) = Bp, ) a) By definition of 6; Pn pik a) = fr vd G7) Hence, SUgh. xa) = 818g. 9. @) = SUP. +s Pe @ by GH) Fir. nd by G7 Thus we have proved (3.5) for x = va By induction, (3.5) is true for all strings x. The other part t.e. the ‘only if part), can be proved similarly. and so (3.4) is established. Now. x € T(M) if and only if &(g, 2x) contains a state of F. By G4). Bide x) Contains a state of F if and only if 54g’ x) is in F’, Hence. x € TM) if and only if x = TUM’). This proves that DFA M' accepts LU Note: In the constriction of a deterministic finite automaton M, equivalent to 2 given nondeterministie automaton , the only difficult par is the construction of 6” for M;. By definition, Fg --- who) = US. a) So we have to apply 6 to (g,, a) for each f= 1. 2 .... k and take their union to get 6g: ».- gah ad. ‘When 6 for M is given in terms of a state table. the construction is simpler. 5iqi, a) is given by the row corresponding to q and the column corresponding to a. To construct 5'(q) «.. qh a). consider the states appearing in the rows corresponding 10 gy...» gi. and the column corresponding to a. These states constitute 3g, ... gal a Note: We write 3” as 6 itself when there is no ambiguity. We also mark the initial state with + and the final state with a circle in the state table. Bae Construct a determin: M = (dy, ai)- (0. 1}. & an tao) where 5 is defined by its state table (see Table 3.2), tic automaton equivalent t0Computer Science TABLE 3.2. Siate Table for Example 36 Solution For the deterministic automaton My (the states are subsets of (go. qh ie. O. fgoh (ao. gib [aik: i) {gol is the initial state: ii) [aol and [qo. qi] are the final states as these are the only states containing gy: and iv) Bis defined by the state table given by Table 3.3 TABLE 3.3. State Table of My for Example 36 Stator ° 1 6 0 9 la (asl fl la fal las ai te ad Ide gd Mad “The states gy and q, appear in the rows corresponding t0 qo and q, and the column corresponding to 0. So. S({qy. aib 9) = (ao. a1) When Mf has m states, the corresponding finite automaton has 2" states However, we need not construct 5 for all these 2" states, but only for those states that are reachable from [go]. This is because our interest is only in constructing M accepting T(M). So, we start the construction of 6 for (gol. We continue by considering only the states appearing earlier under the input columns and constructing 8 for such states. We halt when no more new states appear under the input columns. EXAMPLE 3.7 Find a deterministic acceptor equivalent to M {oe Gi aah: {ay bbs 8 dos Aa) where 6 is as given by Table 3.4 TABLE 3.4. Stale Table for Example 37 Staten aChapter 3: The Theory of Automata _@ 83 Solution ‘The deterministic automaton M, equivalent to M is defined as follows: M. = 2% {ab} & [aah where F = {la2h (o- 92h far. a2) [40 a1. G2) We start the construction by considering [qo] first. We get [q:] and (qo. 4s]. Thon we construct for [qs] and {go. 41}. (gr. 43] is @ new state appearing under the inpat columns. After constructing 6 for [qy, qa}. we do not get any new states and so we terminate the construction af 8. The state table is given by Table 3.5, TABLE 3.5 State Table of M. for Example 37 ‘State @ > Tad (a. a) fea (eal 9 fas. a (a. a fan a) (ad fay. ad Bes) Construct @ deterministic finite automaton equivalent to M = (haw ay 93 a2h (0. 1D). & gos (gob) where 8 is given by Table 3.6. TABLE 3.6 State Table for Example 28 Sato a > sae a % @ e @ a a @ ® @ a Solution Let Q = {go 91, 95 42)- Then the deterministic automaton IM, equivalent to M is given by M.= 22 {a bb. & taol FY where F consists of: Lash (ao. ash Uae Gah Cae ash ba 91. Goh ao 9 Gels Wie toe a2) ane Los i. aes 95) and where d is defined by the state table given by Table 3.7.84 9 Theory of Computer Science TABLE 3.7 State Table of M; for Example 3.8 ‘Staiel cs} tas. ad (a5. ay a3) las gy. 03} 9. 4 a) la. Gy.) fas. a Ge {aoa Gaal 3.8 MEALY AND MOORE MODELS 3.8.1 FINITE AUTOMATA WITH OUTPUTS The finite automata which we considered in the earlier seetions have binary output. ie. either they accept the string or they do not accept the string. This acceptability was decided on the basis of reachability of the final state by the initial state, Now, we remove this restriction and consider the model where the outputs can be chosen from some other alphabet. The value of the output function Zit) in the most general case is a function of the present state g(#) and the present input x11), i Zany = Hain x0) where A is called the output funetion. This generalized model is usually called the Mealy machine. If the output funetion Zi#) depends only on the present state and is independent of the current input, the output function may be written as Zit) = Het) This restricted model is called the Moore machine. I is more convenient to use Moore machine in automata theory. We now give the most general definitions ‘of these machines, Defi jon 3.8 A Moore machine is a six-tuple (Q. 5, 4. & A, qo). where (i) is a finite set of states: fii) E is the input alphaber: (iii) 4 is the output alphaher: Liv) 8 is the transition function E x Q into Q: (v) Ais the output function mapping Q into A; and (si) go is the initia stare Definition 3.9 A Mealy machine is a six-tuple (Q, EA. 6. A go), where all the symbols except 4 have the same meaning as in the Moore machine, 2. is the output funetion mapping © x Q into 3. For example. Table 3.8 describes a Moore machine. The initial state qy is marked with an arrow. The table defines 5 and 2.Chapter 3: The Theory of Automata @ 85 TABLE 3.8 A Moore Machine Present stato Naw! state 6 Output For the input string O111, the ransition of states is given by gq > qs > 9 > 4 > 42 The output string is 00010, For the input string A, the output is Ago) = 0. Transition Table 3.9 describes a Mealy machine TABLE 3.9 A Mealy Machine Present state wate ipa aa o e o % 1 on o Note: For the input sting 0011. the transition of states is given by gh —¥ gs ga ds and the ouput string is 0100. In the case of a Mealy machine. ‘we get an ourput only’ on the application of an input symbol. So for the input string A the output is only A. It may be observed that in the case of a Moore machine. we get Algj) for the input string A. Remark A finite automaton can be converted into a Moore machine by introducing A = (0. 1) and defining A(q) = 1 if g ¢ F and Ag) = 0 if a2F. For a Moore machine if the input string is of length n, the output string is of length m + L. The first output is A(qo) for all output strings. In the case of a Mealy machine if the input string is of length m, the output string is also of the same length 12 3.8.2 PROCEDURE FOR TRANSFORMING A MEALY MACHINE INTO A MOORE MACHINE Ws develop procedures for transforming a Mealy machine into a Moore ‘machine and vice versa so that for a given input string the output strings are the same (except for the first symbol) in both the machines.86_& Theory of Computer Science EXAMPLE 3. Consider the Mealy machine deseribed by the transition table given by Table 3.10. Construct a Moore machine which is equivalent to the Mealy machine. TABLE 3.10 Mealy Machine of Example 3.9 Present state Next slate wate a % ° a ° a a ‘ aw eee o Solution At the first stage we develop the procedure so that both machines accept exactly the same set of input sequences. We look into the next state column for any state, say q,, and determine the number of different outputs associated with. qin that column. We split q, into several different states, the number of such states being equal to the number of different outputs associated with q, For example. in this, problem, qi is associated with one output I and q: is associated with two different outputs 0 and 1. Similarly. 5 and qg are associated with the outputs and 0. 1. respectively. So. we split g> into gay and gy.. Similarly, qs is split into gan and gz). Now Table 3.10 can be reconstructed for the new states as given by Table 3.11 TABLE 3.11 State Table for Example 39 Present state Next state Input a wae 94: % 0 a a 1 tw os a 1 we a 1 a ae 1 % ae 1 % The pair of states and outputs in the next state column can be rearranged as given by Table 3.12.Chapter : The Theory of Automata TABLE 3.12, Revised State Table for Example 3.8 Prosent state Next state ouput a % 1 ca 9 . ° m 3 % 4 % % @ ° a, % @ D a a % 1 Table 3.12 gives the Moore machine, Here we observe that the initial state 4 is associated with output 1, This means that with input A we get an output of 1, if the machine starts at state q,. Thus this Moore machine accepts a zero length sequence (null sequence) which is not avcepted by the Mealy machine, To overcome this situation, either we must neglect the response of a Moore machine (0 input A, or we must add a new starting state qj, whose state transitions are identical with those of q, but whose output is 0. So Table 3.12 is transformed 10 Table 3.13. TABLE 3.13. Moore Machine of Example 3.9 Present state Next state Output os From the foregoing procedure it is clear that if we have an m-output, 1- stare Mealy machine. the corresponding m-output Moore machine has no more than nm + 1 states. 3.8.3 PROCEDURE FOR TRANSFORMING A MOORE MACHINE INTO A MEALY MACHINE We modify’ the acceptability of input string by a Moore machine by neglecting the response of the Moore machine to input A. We thus define that Mealy Machine M and Moore Machine M’ are equivalent if for all input strings w, b7 hve) © Ziplos), where b is the output of the Moore machine for its initial state, We give the following result: Let M; = (Q. , A. d A, qy) be a Moore machine, Then the following procedure may be adopted t0 construct an equivalent Mealy machine Ms.88_@ Theory of Computer Science Construction (i) We have to define the output function 2 for the Mealy machine as a function of the present state and the input symbol, We define 2’ by H(q, a) = MS{a, «)) for all states q and input symbols a (ii) The transition function is the Same as that of the given Moore ‘machine. EXAMPLE 3.10 Consteuet « Mealy Machine which is equivalent to the Moore machine given by Table 3.14. TABLE 3.14 Moore Machine of Example 3.10 Present state a % % Solution We must follow the reverse provedure of converting a Mealy machine into a ‘Moore machine. In the case of the Moore machine, for every input symbol we form the pait consisting of the next state and the corresponding output and reconstruct the table for the Mealy Machine. For example. the states q3 and qy in the next state column should be associated with outputs 0 and i, respectively The transition table for the Mealy machine is given by Table 3.15. TABLE 3.15 Mealy Machine of Example 3.10 Prasent state Next state auto Sao att 0 a 1 1 % ° ° a ° ° & ° Note: We can reduce the number of states in any model by considering states with identical transitions. If two states have identical wansitions (i.e. the rows, corresponding t these two states are identical), then we can delete one of ther. EXAMPLE 3.11 Consider the Moore machine described by the transition table given by Table 3.16. Construct the corresponding Meuly machine.Chapter 3: The Theory of Automata 89 TABLE 3.16 Moore Machine of Example 3.11 Next state Cutpot a © % 6 % pa Solution We construct the transition table as in Table 3.17 by associating the output with the transitions, In Table 3.17. the rows corresponding to > and qs are identical. So, we ean deleie one of the two states. ie. gz or qs. We delete gs. Table 3.18 gives the reconstructed table TABLE 3.17 Transition Table for Example 3.11 Present stare Next state TABLE 3.18 Mealy Machine of Example 3.11 Prasant st Noxt state In Table 3.18. we have deleted the qs-row and replaced qs by q2 in the other Consider a Meaiy machine represented by Fig. 3.10, Construct a Moore machine equivalent to this Mealy machine908 Theory of Computer Science Start Wy Fig. 310 Mealy machine of Example 3.12 Solution Let us convert the transition diagram into the transition Table 3.19. For the given problem: q, is not associated with any output; q> is associated with Iwo different outputs Z, and Zs: dy is associated with two different outputs Z, and Zp, Thus we must split g> into qs) and qo» with outputs Z; aud Z,, respectively and qs into qs, and gs> with outputs Z; and Z,, respectively. Table 3.19 may be reconstructed as Table 3.20. TABLE 3.19 Transition Table for Example 3.12 Present state Next state % a TABLE 3.20 Transition Table of Moore Machine for Example 3.12 Present state Next state Output Taso et a a a om ae ae Figure 3.11 gives the transition diagram of the required Moore machine.Chapter 3: The Theory of Automata © 91 Fig. 3.11 Moore machine of Example 3.12 3.9 MINIMIZATION OF FINITE AUTOMATA In this section we construct an automaton with the minimum number of states equivalent to a given automaton Mf. As our interest lies only in strings accepted by M, what really matters is whether a state is a final state or not, We define some relations in Q. Definition 3.10 Two states q, and qy are equivalent (denoted by qy = 42) if both 8{qy, x) and (qx) are final states. oF both of them are nonfinal states for all xe E* As it is difficult to construct &(q,, x) and &{qs, x) for all x € E* (there are an infinite number of strings in E*). we give one more definition Definition 3.11. Two states q, and q; are K-equivalent (k 2 0) if both iq, x) and 4(q>. x) are final states or both nonfinal states for all strings x of length kor less. In pecticular, any two final states are O-equivalent and any two nonfinal states are also O-equivalent We mention some of the properties of these relations. Property 1 The relations we have defined, ic. equivalence and k-equivalenee, ate equivalence relations. i.e. they are reflexive. symmetric and transitive. Property 2 By Theorem 2.1. these induce partitions of Q. These partitions can be denoted by and 7. respectively. The elements of 7; are k-equivalence classes. Property 3 If q, and g: are k-equivalent for all k2 0. then they are equivalent. Property 4 If q; and qs are (k + L)-equivalent. then they are k-equivalent. Property § x, = Myx for some n. (x, denotes the set of equivalence classes under n-equivalence,) The following result is the key to the construction of minimum state automaton, RESULT Two states g and gz are (k + L)-equivalent if (i) they are A-equivalent; (ii) 6(q,, a) and 6(q3 a) are also k-equivalent for every a € E.92S Theory of Computer Science Proof We prove the result by contradiction. Suppose q, and q> are not (c+ Deequivatent. Then there exists a string w= aw, of length k + 1 such that 8(q), cov) is a final state and 5¢qz, aw,) is not a final state (or vice versa; the proof is similar). So 6(8(q), a), w)) is a final state and 8(8(q>, a), wy) is not 4 final state, As w) is a string of length & S(q,, a) and S(qy, a) are not equivalent. This is a contradiction, and hence the result is proved. 1 Using the previous result we can construct the (k + 1)-equivalence classes once the K-equivalence classes are known. 3.9.1 CONSTRUCTION OF MINIMUM AUTOMATON Step 1 (Construction of n). By definition of 0-equivalence, m = {Q?, O89} where Qf is the set of all final states and Q? = Q- 0? Step 2 (Construction of 7j.; from 7). Let Q be any subset in m. If q, and gz are in Qf, they are (K+ 1)-equivalent provided 5(q,, a) and 6(q3, a) are equivalent, Find out whether 5(q), a) and 5(q>, a) are in the same equivalence class in m, for every a ¢ E. Hf so. gi and q: are (k + 1)-equivalent. In this way, Q} is further divided into (k + 1)-equivalence classes. Repeat this for every Qi in to get all the elements of Z,} Step 3. Consiruc Step 4 (Construction of minimum automaton). For the required minimum state automaton, the states are the equivalence classes obtained in step 3. i.e. the elements of #,. The state table is obtained by replacing a state q by the corresponding equivalence class [q) my form = 12... until m= Fy Remark In the sbove construction, the crucial part is the construction of equivalence classes: for, after getting the equivalence classes, the table for ‘minimum automaton is obtained by replacing states by the corresponding equivalence classes. The number of equivalence classes is less than or equal to 10. Consider an equivalence class {] = (gy. az =.» dx} If gy is reached while processing ws € TUM) with (aq, w;) = gi, then 5(g, w:) © F. So, Sigg 3) F for i= 2. ....& Thus we see that 9}. 7 = 2, .... kis reached on processing some w € T(M) iff qy is reached on processing w, Le. q, of [qi] can play the role of q».. qs. The above argument explains why we replace a state by the corresponding equivalence class. Note: The construction of m, 7, m ete. is easy when the transition table is given, m = (0°, 09}. where QP = F and Q? = Q ~ F. The subsets in 1% are obtained by further partitioning the subsets of a. If gi. gz © QP, consider the states in each a-column, where a € E corresponding to q, and 4» If they are in the same subset of m, qy and qy are I-equivalent. If the states under some a-column are in different subsets of %, then g, and qy are not L-equivalent. In general, ( + 1)-equivalent states are obtained by applying the above method for q and qs in OF__Ghapter 3: The Theory of Automata _@ 93 Ez Construct a minimum state automaton equivalent to the finite automaton described by Fig. 3.12. Fig. 3.12 Finite automaton of Example 3.13, Solution It will be easier if we construct the transition table as shown in Table 3.21 TABLE 3.21 Transition Table for Example 3.13 By applying step 1, we get OP = F = {qh So. Ry = [142]. {do i 9% Ge Os Ie ah}942 Theory of Computer Science The {q:) in m cannot be further panitioned. So, Q', = {q}. Consider qo and 4a; € 8. The entries under the O-cclumn corresponding 10 gp and qj; are 4) and qg: they lie in O°. The entries under the 1-column are gs and qo. q> € 0? and qe € Q2. Therefore, qo and q, are not L-equivalent. Similarly. qo is not I-equivalent to q:. gs and 4) Now, consider qy and qs. The entries under the O-column are q; and 4 Both are in QS. The entries under the 1-column are qs, qs- So qs and qo are L-equivalent. Similarly. gq is I-equivalent to gg. {qo gu do) is a subset in 7, So, Os = ta da ah: Repeat the construction by considering q, and any one of the states 45. gs. dr- Now, gy is not L-equivalent to qs or gs but I-equivalent to qr. Hence, 0% (4, qr}. The elements left over in Q! are qs and gs. By considering the entries under the O-column and the I-column, we see that gs and ge ate equivalent, So Qs = {gz 4s). Therefore. = Uda}. (do de Gols (aie rhe (gs ash} ‘The’ (q3} is also in zt; as it cannot be further partitioned. Now. the entries under the O-column corresponding 10 qo and g, are g; and qr, and these lie in the same equivalence class in m, The entries under the L-column are qs, qs So qy and qy are 2-eqnivalent, But qy and gy are not 2-equivatent. Hence. {ao 40 go} is partitioned into (qo, 94} and (qg)- q, and qr are 2-equivalent gz and gs ate also 2-equivalent. Thus. = {{qo}. {do dal. {gels (au ah fas. qsh}- gy and gy are 3-equivalent. The q, and qr are J-equivalent, Also, gz and qs are 3-equivalent. Therefore. Y Udiv Gabe tgols Lane arb fare ash AS Ms = my My gives us the equivalence classes. the minimum state automaton m= Ly = 410.1). 6 4%. FY where Q°= (ash [dos dah lash Lave ash fas 5) qo (do ask F’= (aad and 6 is defined by Table 3.22. TABLE 3.22 Transition Table of Minimum State Automaton for Example 3.13 ‘Stators 1 (as. ad fa. od fe. ad fa. a (eel taal fal fas, aa tel fas. a8) (al laa teal ta (> a)— Chapter 3: The Theory of Automata _& 95 Note: The transition diagram for the minimum state automaton is described by Fig. 2.13, The states qo and q, are identitied and treated as one state (So also are qj, gy and qs, qx) But the transitions in both the diagrams (ie. Figs. 3.12 and 3.13) are the same. If there is an arrow from q; to q; with label @. then there is an arow from [qi] 1 {gd with the same label in the diagram for minimum state automaton. Symbolically, if Bq. a) = gp then Sg. a) = tg). Fig. 3.13 Minimum state automaton of Example 3.13. Construct the minimum stare automaton equivalent to the transition diagram given by Fig. 3.14. Fig. 3.14 Finite automaton of Example 3.14 Solution We construct the transition table as given by Table 3.2:rv of Computer Science TABLE 3.23 Transition Table for Example 3.14 Sia 2 ° ars a rs a & a By & a ® & % % & & & % a a & a a 4 Since there is only one final state qz, Q\' = {43}, Q? = Q - QP. Hence, m= (ash. (qo 4 42 46 95 Ae Jr}}- AS {43} cannot be partitioned further, Qi = {ga}. Now qo is L-equivalent to 9}, 45, qo But mot {0 qs, ga. gr. and so 2 = (do. Mi» 45. Go): 42 18 L-equivalent to qy Hence, Q’; = (42 44}. The only element remaining in Q! is qj. Therefore, Q's = {q;}- Thus, m= (lash {do ds 9 dol (a> debs tard} QF = Lash gp is 2equivalent to gq but not t0 q, oF ge. So. 23 = (a a6) As qi is 2-equivalent tog, OF = (av ash As qz is 2-equivalent t0 du, QP = lands QF = ta} Thus, = {Las}, (do. d}- {a as)+ {a daly (art) QF = tas) As qo is equivalent t0 ge, (Go. Go) As q: is 3-equivalent to qs, {qu gst AS qz is 3-equivalent to 44, Qi = {qx 44). (an) Therefore, Aas}. {go debs (die as), (a> ashe tar}t 7M, My gives us the equivalence classes, the minimum state automaton M’ =(Q, (ab). 8 gi. FY)Chapter 3: The Theory of Automata 8 97 Q'= (lash [ao teh Cav ash fas. ah Cal) = [go oh F= Kas and 5° is defined by Table 3.24, TABLE 3.24 Transition Table of Minimum State Automaton for Example 3.14 ‘Stater= a (oe. a Gan ad (eo. al ta. ad toe ad te ad tt teal eal tal (a (ao. ae) k [ao ae} (al Note: ‘The transition diagram for M’ is given by Fig. 3.15, oS Cena lead tol A NE Fig, 3.18 Minimum state automaton of Example 3.14 3.10 SUPPLEMENTARY EXAMPLES EZ Construct a DFA equivalent to the NDFA M whose transition diagram is given by Fig. 3.16. Fig. 3.16 NOFA of Example 3.16.98 _G Theory of Computer Science Solution The transition table of Mf is given by Tuble 3.25. TABLE 3.25 Transition Table for Example 3.18 g = % a * & ao * a ~ ®@ - ® = = For the equivalent DFA: i) The states are subsets of Q = ty. a1. a> 92 Gab Gi) [go] is the initial stae. Git) The subsets of Q containing qs or gs ste the final states (iv) dis defined by Table 3.26. We start from [qo] and construct 6 only for those states reachable trom {qa} (as in Example 3.8 TABLE 3.26 Transition Table of OFA for Example 3.15 ‘State @ > (al lead (a. a) (ae. aa) fe. ad) fae. ad (oud Construct a DEA equivalent to an NDFA whose transition table is defined by Table 3.27. TABLE 3.27 Transition Table of NOFA for Example 3.16 ‘State e ° % a Ge de a 4 % % % % Solution Let M be the DEA defined by M = (hie 5), fa, B), & [gal F)where F is the set of all subsets of {qo 1. qs. qa} containing ¢y. 6 is defined by Table 3.28, TABLE 3.28 Transition Table of OFA for Example 3.16 ‘State Ete sda Construct a DFA accepting all strings w over {0, 1} such that the number of I's in w is 3 mod 4 Solution Let M be the required NDFA. As the condition on strings of TM) does not at all involve 0, we ean assume that Mf does not change state on input O. If appears in w (4k + 3) times. M can come back to the initial stae, after reading 41's and to a final state after reading 3 1's. The required DFA is given by Fi ° ‘, ar ‘eT ayo O_o»: KS 7 ey Fig. 3.17 DFA of Example 3.17. Ura Construct 2 DFA accepting all strings over (a, b) ending in ab. Solution We require two transitions for accepting the string ab. If the symbol b is processed after aa or ba, then also we end in ad. So we can have states for100 _Theory of Computer Science remembering aa, ab, ba, bb. The state corresponding to ab can be the final state in our DFA. Keeping these in mind we construct the required DFA. Its transition diagram is described by Fig. 3.18. Fig. 318 DFA of Example 3.18. EXAMPLE 3.19 Find T(M) for the DEA M described by Fig. 3.19 oe A A _ Solution Tif) = {w © {a, b}* | ends in the substring ab} Note: If we apply the minimization algorithm to the DEA in Example 3.18, we get the DFA as in Example 3.19. (The student is advised 10 check.) EZUaero Construct a minimum state automaton equivalent to an automaton whose transition table is defined by Table 3.29.Chapter 3: The Theory of Automata @ 101 TABLE 3.29. DFA of Example 3.20 sae ppesss BRReae "| Solution OP = lash. OP = (ar diy des a dab 80. % = Ads) (ay av 42 I Gab) Q? vannot be panitioned further. So {qs} € 7)- Consider Q2. qy is equivalent to gj. qy and gy But q is not equivalent to gs since S(gq, b) = gz and 8(q3. B) = Lash Hence, OF = (ah. OL = ta a a a). 0} = Las) Therefore, = (dss {do ga}. fast qo is Qequivalent to g, but not 2equivalent to gy oF gs Hence, ap. 9) © qy and gz are not 2-equivalent. ‘Therefore. Ry = thas}- Aashe (ao ashe (9i}> (agald As qo is not 3-equivalent to 4, (go. qu} is further partitioned into {qx} and a4) So, = (dol. (aib fash Lash (ass tas Hence the minimum state automaton Mf is the same as the given M. EXAMPLE 3.21 Construct a minimum state automaton equivalent to a DEA whose transition table is defined by Table 3.30.102_&_Theory of Computer Science Solution OP = tgs. gabe OP = L4oe div do 5 dor 47} Ty= (gs Gals {0 div I» se Gos rh) 45 18 L-equivalent to qs. So, (qs. dah © 7 go is not Lequivalent to g.. 4 ds but gy is equivalent 10 go, Hence {qu qs) € %- qi is L-equivalent to q; but not L-equivalent 10 G5 4 OF Gr So, {qe ao} My gs is not Lequivalent to q, but 0 Hence, $0. {ds ar} € Fi ™ = Udy dade dor do) (ae aah as eh) gz is Qequivalent to qs. So. {qx qa} & Me go is not 2-equivalent to qe So. do)+ {a6} © A> gx is Qequivalent t0 qo. So. {qy. gz} © gs is Zequivalent t0 gr. So. (qs. gs) © 7 Hence, = Ugx gabe Aol (doh tai a2) tas ahh 4g; 8 Bequivalent to 4: q, is equivalent to gs and qe is F-equivalent t0 ¢ Hence, (dole tate debe tas as)- dss arhs dob the minimum state automaton is ts As M’ = (Q% (a. b}. &% Lagu (las auld) where 8’ is defined by Table 3.31 TABLE 3.31 Transition Table of DFA for Example 321 ‘State a ° lad fas. aa) las. al lan ad las. ad las. al te, aa las a) {cal (as. a1 Ia. aa fal cy (ae fal i i | 5 f i tChapter The Theory of Automata 103 SELF-TEST Study the automaton given in Fig. 3.20 and choose the correct answers to Questions 1-8: Fig. 3.20 Automaton for Questions 1-5 1 Misa (a) nondeterministic automaton (b) deterministic automaton accepting {0. 1}* (6) deterministic automaton accepting all strings over (0. 1} having 3a O's and 3n V's. mn 2 (€) deterministic automaton 2. M accepts (@) 01110 () 10001 () 01010 (@ it 3. TUM) is equal to (ay (0" 1" | m,n 2 O} (b) {0 P*]m. m2 1} (6) foe] has L11 as a substring) (2) fue }w has u's. 2 4. If gp is also made a final state, then Mf accepts (2) 01110 and 01100 (b) 10001 and 10000 (e) 0110 but not 0111101 (a) 0". m2 1 but not Pn 2 S$. If gz is also made « final state, then TUM) is equal t0 (2) {0° 1 fm, 1 20} Y (OP I" | ma, 2 2 OF () (OE fom BAPE OPM | mB HY (c) few has 111 as a substting or LI as a substring} (@) tor] the number of 1's in w is divisible by 2 or 3} Study the automaton given in Fig. 321 and state whether the Statements 6-15 are true or false: at os Es a bBo 6 Fig. 3.21 Automaton for Statements 6-15.104 52 _Theory of Computer Science 6 7. 8. 9. 10. n 2 2B. 4. 48. 31 32 33 34 3S 36 BT M is a nondeterministic automaton, S(qy. 1) is defined. 0100111 is accepted by M. 010101010 is not accepted by M. 6(qy. 01001) = {ar} (qo. 011000) = (qo. 94. a2} 6(q2, ¥) = gs for any string w € {0. 1}*. 6(q,. 11001) # 0 TOM) = (w|w = x00y, where x, y € (0. 1)*}. A string having an even number of 0's is accepted by M. EXERCISES For the finite state machine M described hy Table 3.1. find the strings. among the following strings which are accepted by M: (a) 101101, (b) TIL, (©) 000000. For the transition system M described by Fig. 3.8, obtain the sequence of states for the input sequence 000101. Also. find an input sequence not accepted by M. Test whether 110011 and 110110 are accepted by the transition system described by Fig. 36. Let M = (Q, E. & qo F) be a finite automaton. Let R be a relation in @ defined by q\Rqy if 61g}. a) = (gz, a) for all ae E Ts Ran equivalence relation? Construct @ nondeterministic finite automaton accepting {ab, ba}, and use it to find a deterministic automaton accepting the same set. Construct @ nondeterministic finite automaton accepting the set of all strings over {a, 6} ending in aba, Use it to construct a DFA accepting the same set of strings The transition table of a nondeterministic finite automaton M is defined by Table 3.32. Construct a deterministic finite automaton equivalent to M. TABLE 3.32 Transition Table for Exercise 3.7 0 aa % 3 ooChapter Theory of Automata 105 3.8 Construct a DFA equivalent to the NDFA described by Fig. 3.8. 39M = (1g. gy as), (0. 1}. & qu. (gah) is a nondeterministic finite automaton, where 6 is given by 6. = {ay ash Sig. Y= fait 5q> O= (a. a2} 6a N= OH 6(gs, 0) = (42). Sx Y= {a ah Construct an equivalent DFA. 3.10 Construct a transition system which can accept strings over the alphabet 4, b, ... containing either cat or rat 3.11 Construct a Mealy machine which is equivalent to the Moore machine defined by Table 3.33, TABLE 3.33. Moore Machine of Exercise 3.11 Next state Output at e 1 & ° a 1 % 1 3.12, Construct a Moore machine equivalent to the Mealy machine M defined by Table 3.34, TABLE 3.34 Mealy Machine of Exercise 3.12 Present state ext stare aie “aie “a a % 0 % % % 1 % a % 1 a % a 1 343 Construct a Mealy machine which can output EVEN. ODD according as the total number of I's encountered is even or odd. The input symbols are O and 1 3.14 Construct a minimum state automaton equivalent to a given automaton M whose transition table is defined by Table 3.35.106 8 Theory of Computer S TABLE 3.35 Finite Automaton of Exercise 3.14 State ‘Input esasaes lama > 3.15 Construct a minimum state automaton equivalent to the DFA described by Fig. 3.18. Compare it with the DFA described by Fig. 3.19. 3.16 Construct a minimum state automaton equivalent to the DFA described by Fig. 3.22. on Fig. 3.22 DFA of Exercise 3.16.