Toward a Uniform Approach to the Unfolding of Nets

In this paper we introduce the notion of spread net. Spread nets are (safe) Petri nets equipped with vector clocks on places and with ticking functions on transitions, and are such that vector clocks are consistent with the ticking of transitions. Such nets generalize previous families of nets like unfoldings, merged processes and trellis processes, and can thus be used to represent runs of a net in a true concurrency semantics through an operation called the spreading of a net. By contrast with previous constructions, which may identify conflicts, spread nets allow loops in time

💡 Research Summary

The paper introduces “spread nets,” a novel class of safe Petri nets equipped with vector clocks on places and ticking functions on transitions. A spread net augments the traditional token game with additional information that records the local execution history of each component of a system. This is achieved by defining multi‑clock nets (mc‑nets) that view a safe net as a collection of synchronising automata, each identified by a mapping ν from places to the initial marking. The authors then construct ticking domains: for an alphabet A, an equivalence relation ∼Eq on A* generated by a set of equations Eq (e.g., s = t, suvs = s) yields a quotient A_Eq = A* / ∼Eq. Each component i receives its own ticking domain A_i, and the Cartesian product of all A_i forms a vector‑clock domain A = ×_i A_i. A vector clock is a tuple of equivalence classes, one per component, representing the component’s local run up to the current point.

Transitions are associated with ticking functions δ that map a vector clock before firing to a new vector clock after firing. A transition t is enabled not only when its preset places contain tokens, but also when the current vector clock matches the ticking function’s requirements. Firing t consumes tokens, produces tokens, and updates the vector clock according to δ(t). Because the vector clock can return to a previously seen equivalence class, the resulting spread net need not be a directed acyclic graph; loops in time are allowed, which distinguishes this approach from classic unfoldings.

The central operation, called “spreading,” takes an mc‑net as input and constructs its corresponding spread net by iteratively applying transitions and updating vector clocks. The authors prove that spreading preserves important algebraic properties: it is compositional, respects morphisms between nets, and yields a net that is a homomorphic image of the original. Moreover, they show that traditional unfoldings, merged processes, and trellis processes are special cases of spread nets obtained by selecting particular ticking domains and ticking functions. In categorical terms, safe nets and their morphisms form the category Safe, mc‑nets form a subcategory MCN, and spreading defines a functor from MCN to a category of spread nets that preserves structure.

By unifying time and conflict handling through the same mechanism (vector clocks and ticking functions), spread nets provide a compact representation of concurrent executions while allowing cycles in the temporal dimension. This addresses the state‑space explosion problem inherent in classic unfoldings, as the equivalence classes can collapse infinitely many distinct interleavings into a finite representation. The paper also discusses potential applications in model checking, verification of concurrent systems, and real‑time analysis, and outlines future work on automated generation of ticking domains, efficient spreading algorithms, and tool integration.

In summary, the authors present a mathematically rigorous framework that generalizes existing net‑unfolding techniques, offers a uniform treatment of causality and conflict, and opens new avenues for scalable analysis of concurrent systems.