Team equivalences for finite-state machines with silent moves
Introduction
By finite-state machine (FSM, for short) we mean a simple type of finite Petri net [15], [29], [33] whose transitions have singleton pre-set and singleton, or empty, post-set. The name originates from the fact that a net of this kind is isomorphic to a nondeterministic finite automaton (NFA), usually called a finite-state machine as well. However, semantically, our FSMs are richer than NFAs because, as their initial marking may be not a singleton, these nets can also exhibit concurrent behavior, while NFAs are strictly sequential. FSMs are also very similar to finite-state, labeled transition systems (LTSs, for short) [22], a class of models that are suitable for describing sequential, nondeterministic systems, and are also widely used as a semantic model for process algebras (see, e.g., [14]). On this class of models, there is widespread agreement that a very natural and convenient equivalence relation is bisimulation equivalence [28], [25]. If the LTS contains silent transitions, i.e., transitions labeled by the invisible action τ, then Milner proposed weak bisimulation equivalence [25] as a natural extension of bisimulation equivalence to this setting. However, van Glabbeek and Weijland in [12] argued that weak bisimilarity does not completely respect the branching structure of processes and so they proposed branching bisimulation equivalence as a suitable equivalence in the presence of silent moves.
In [16] we defined a new truly-concurrent equivalence relation over FSMs (without silent moves), called team equivalence, that can be computed in a distributed manner, without resorting to a global model of the overall behavior of the analyzed marked net. Since an FSM is so similar to an LTS, the basic idea we started with was to define bisimulation equivalence directly over the set of places of the unmarked net. The advantage is that bisimulation equivalence is a relation on places, rather than on markings (as it is customary for Petri nets), and so much more easily computable; more precisely, if m is the number of net transitions and n is the number of places, checking whether two places are bisimilar can be done in time, by adapting the optimal algorithm in [30] for standard bisimulation on LTSs. After the bisimulation equivalence over the set of places has been computed, we can define, in a purely structural way, that two markings and are team equivalent if they have the same cardinality, say , and there is a bisimulation-preserving, bijective mapping between the two markings, so that each of the k pairs of places , with and , is such that and are bisimilar. Team equivalence is a truly concurrent behavioral equivalence as it is sensitive to the size of the distributed state; as a matter of fact, it relates markings of the same size, only. The name team equivalence reminds us that two distributed systems, composed of a team of non-cooperating, sequential processes, are equivalent if it is possible to match each sequential component of the first system with one bisimulation-equivalent, sequential component of the other one, as in any sports where two competing (distributed) teams have the same number of (sequential) players. Once bisimilarity on places has been computed, checking whether two markings of size k are team equivalent can be computed in time (or , cf. Remark 1).
Note that to check whether two markings are team equivalent we need not to construct an LTS describing the global behavior of the whole system, but only to find a suitable, bisimulation-preserving match among the local, sequential states (i.e., the elements of the markings). Nonetheless, we proved that team equivalence is coherent with the global behavior of the net. More precisely, we showed in [16] that team equivalence is finer than interleaving bisimilarity, actually it coincides with strong place bisimilarity [1] and so it respects the causal semantics of nets.
The main goal of this paper is to extend our previous definition of (strong) bisimulation on places (and of team equivalence) to FSMs with silent moves, taking inspiration from Milner's weak bisimulation [25], [14] and van Glabbeek and Weijland's branching bisimulation [12], [14]. Therefore, we first define weak bisimulation on places and its associated weak team equivalence, together with the variants requiring the so-called rootedness condition (i.e., an initial silent move can be matched only by a nonempty sequence of silent moves, as in rooted weak bisimilarity [25], [14]). Then, we define branching bisimulation on places and its associated branching team equivalence, together with the variants requiring the rootedness condition (i.e., the first move is to be matched strongly, as in rooted branching bisimilarity [12], [14]). The originality of our proposal is not in the technical definition of (rooted) weak/branching bisimulations on places (which are, indeed, almost identical to the original ones on LTSs), rather on the fact that these relations are defined over the places of an unmarked net, rather than on the reachable markings of a marked net. Moreover, the main originality of our proposal is in the definition of (rooted) weak/branching team equivalences; these equivalences are all computed in a structural way, without building a model of the global behavior. Nonetheless, we will prove that these are coherent with the global behavior; in particular, they are finer than the corresponding (rooted) weak/branching interleaving bisimulation equivalences over FSMs (see Section 2 for details), which are equivalences relations defined over the net markings. The key feature common to all the new team equivalences we present in this paper, is that, contrary to the weak/branching interleaving bisimulation equivalences, to a silent move of a single sequential component of the marking , the marking may reply only with a (possibly empty) sequence of silent transitions which are local to one of its sequential components.
In [15] we proved that the class of FSMs can be “alphabetized” by means of the process algebra CFM: not only the net semantics of each CFM term is an FSM, but also, given a FSM N, we can single out a CFM term such that its net semantics is an FSM isomorphic to N. This means that we can define team equivalences also over CFM process terms. CFM is a simple process algebra: it is actually a slight extension to finite-state CCS [25] and a subcalculus of both regular CCS and BPP [14].
Based on our previous work [17], where we provided a sound and complete axiomatization for (strong) team equivalence over CFM, the goals of the second part of this paper are three: we want (i) to prove that rooted weak/branching team equivalence is a congruence for the CFM operators, (ii) to study the algebraic properties of these equivalences and, finally, (iii) to provide them with a sound and complete, finite axiomatization for CFM. These axiomatizations are not too surprising: it is enough to add to (a slightly revised version of) the finite axiomatization of rooted weak/branching bisimulation equivalence for finite-state CCS [26], [11], three axioms for parallel composition stating that it is associative, commutative and with 0 as neutral element. However, the technical treatment is different (and somehow simpler) than [26], [11], as we base our axiomatization on guarded process constants (e.g., ) rather than on the recursion construct (with possible unguarded variables; e.g., ). To the best of our knowledge, these are the first axiomatizations of a truly concurrent behavioral equivalence in the presence of silent moves.
The paper is organized as follows. Section 2 introduces the basic definitions about finite-state machines and some well-known behavioral equivalences on them: (rooted) weak interleaving bisimilarity and (rooted) branching interleaving bisimilarity. Section 3 copes with the distributed equivalence checking problem for (rooted) weak equivalence; first, (rooted) weak bisimulation over places of an unmarked net is defined; then, (rooted) weak team equivalence is introduced and some examples are presented, together with a proof that it is finer than (rooted) weak interleaving bisimilarity; moreover, the minimization of an FSM w.r.t. weak bisimilarity on places is defined. Section 4 copes with the similar distributed equivalence checking problem for (rooted) branching bisimilarity: we define first (rooted) branching bisimilarity on places, then (rooted) branching team equivalence and its minimized net. Section 5 introduces the process algebra CFM, its syntax, its net semantics and recalls the so-called representability theorem from [15]. Section 6 shows that rooted weak/branching team equivalences are congruences for the CFM operators and studies their algebraic properties. Section 7 presents the finite axiomatizations of rooted weak/branching team equivalences for CFM, proving that they are sound and complete. Finally, Section 8 discusses some related literature and future research.
Section snippets
Basic definitions and behavioral equivalences
Definition 1 (Multiset) Let be the set of natural numbers. Given a finite set S, a multiset over S is a function . The support set of m is the set . The set of all multisets over S, denoted by , is ranged over by m, possibly indexed. We write if . The multiplicity of s in m is given by the number . The size of m, denoted by , is the number , i.e., the total number of its elements. A multiset m such that is called empty and is denoted by θ. We
A distributed approach to weak equivalence checking
We recall the definition of (strong) bisimulation on places for unmarked FSMs, originally introduced in [16], as it will be useful in the following. In this definition (and in the following ones), the markings and can only be either the empty marking θ or a single place, because of the shape of FSM transitions.
Definition 11 (Bisimulation on places) Let be an FSM. A bisimulation on places is a relation such that if then for all such that , such that and
A distributed approach to branching equivalence checking
In [12], van Glabbeek and Weijland argued that weak bisimilarity ≈ is not completely respecting the timing of choices (the so-called branching structure of systems). For instance, consider the two nets in Fig. 8. A weak bisimulation is , , hence (actually, ). However, in the net in (a), in each computation the choice between b and c is made after the a-labeled transition, while in the net in (b) there is a computation where c is already discarded by
CFM: syntax and net semantics
Now we define the process algebra CFM [15] (where CFM is the acronym of Concurrent Finite-state Machines) that truly represents FSMs.
Congruence and algebraic properties
Thanks to the theorems of the previous section, we can transfer the definitions of the various team-based bisimulation equivalences from FSM nets to CFM process terms in a simple way.
Definition 19 Two CFM processes p and q are (strong) team bisimilar, denoted by , if, by considering the (union of the) nets and , holds. In the same way, we can define all the other team equivalences; for instance, if and if . □
Of course, for sequential
Axiomatizations
Now we provide a sound and (ground-)complete, finite axiomatization of rooted weak/branching team equivalence over observationally guarded CFM processes. For simplicity's sake, the syntactic definition of open CFM is given with only one syntactic category, but each ground instantiation of an axiom must respect the syntactic definition of CFM given in Section 5; this means that we can write the axiom , but it is invalid to instantiate it to because these
Conclusion
Finite-state machines with silent moves have been equipped with simple, efficiently decidable, truly-concurrent behavioral semantics. Indeed, weak (or branching) team equivalence seems the most natural, intuitive and simple extension of LTS weak (or branching) bisimulation equivalence to FSMs with silent moves; it also has a very low complexity, actually the lowest one for FMSs with silent moves. More precisely, weak bisimilarity on places can be checked in time (where n is the
Declaration of Competing Interest
The author declares that he has no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
The anonymous referees are thanked for their detailed comments and suggestions.
References (36)
Branching bisimilarity is an equivalence indeed!
Inf. Process. Lett.
(1996)- et al.
Sequential and concurrent behavior in Petri net theory
Theor. Comput. Sci.
(1987) Verification of finite-state machines: a distributed approach
J. Log. Algebraic Methods Program.
(2018)A complete inference systems for a class of regular behaviors
J. Comput. Syst. Sci.
(1984)A complete axiomatisation for observational congruence of finite-state behaviours
Inf. Comput.
(1989)- et al.
Generalizing the Paige-Tarjan algorithm by abstract interpretation
Inf. Comput.
(2008) Bisimulation and action refinement
Theor. Comput. Sci.
(1993)- et al.
Strong bisimilarity on nets revisited
- et al.
Place bisimulations in Petri nets
- et al.
A ground-complete axiomatisation of finite-state processes in a generic process algebra
Math. Struct. Comput. Sci.
(2008)
Concurrent bisimulations in Petri nets
Acta Inform.
Partial ordering descriptions and observations of nondeterministic concurrent systems
Algorithm 97: shortest path
Commun. ACM
Petri net models for algebraic theories of concurrency
Equivalence notions for concurrent systems and refinement of actions
A complete axiomatization for branching bisimulation congruence of finite-state behaviours
Branching time and abstraction in bisimulation semantics
J. ACM
Structure preserving bisimilarity - supporting an operational Petri net semantics of CCSP
Cited by (7)
A study on team bisimulation and H-team bisimulation for BPP nets
2022, Theoretical Computer ScienceCitation Excerpt :A first step in this direction is [30].
Branching Place Bisimilarity
2023, arXivDistributed Non-Interference
2023, arXivBranching Place Bisimilarity: A Decidable Behavioral Equivalence for Finite Petri Nets with Silent Moves
2021, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)