Stochastic Well-Structured Transition Systems

• Stochastic well-structured transition systems are defined as a quintuple (S, ≼, →, |·|, Pr) that extends WSTS with probabilistic transitions and a size function, unifying various distributed stochastic models.
• They ensure structural monotonicity and decidability of upward reachability using a simulation lemma and the finite-basis property of well-quasi-orders.
• SWSTS precisely capture computational classes by enforcing polynomial termination bounds, characterizing BPP in uniform systems and symmetric BPL in log-space models.

Stochastic well-structured transition systems (SWSTS) generalize the established framework of well-structured transition systems (WSTS) by introducing size functions and a probabilistic scheduler constrained by inverse-polynomial lower bounds. This extension yields a unified model that encompasses population protocols, chemical reaction networks, various synchronous gossip/token-passing protocols, as well as their augmentations with order or equivalence oracles. SWSTS offer both structural monotonicity and a probabilistic guarantee per transition and admit exact characterizations of their computational power in terms of classical randomized complexity classes—specifically, BPP for equipped models and the symmetric languages in BPL for unaugmented models (Aspnes, 24 Dec 2025).

1. Formal Structure and Definitions

An SWSTS is a quintuple $(S,\preceq,\to,|\cdot|,\Pr)$, building on WSTS by enriching the transition relation with probabilistic structure:

• State space $S$ equipped with a well-quasi-order (wqo) $\preceq$.
• Transition relation $\to$ with the monotonicity property: for any $s \to s'$ and $s \preceq t$, there exists $t'$ such that $t \to t'$ and $s' \preceq t'$.
• Weight function $|\cdot|: S \rightarrow \mathbb{N}$ assigns a size to each configuration.
• Transition probability function $\Pr: S \times S \rightarrow [0,1]$ such that for every $s \in S$, $\sum_{s': s \to s'} \Pr(s \to s') = 1$.

The defining polynomial-probability condition requires that for each $s \to s'$, there exist $k \ge 0, c > 0$ such that for any $t \succeq s$,

$$\sum_{\substack{t \to t' \ t' \succeq s'}} \Pr(t \to t') \geq c \cdot |t|^{-k}.$$

A SWSTS is said to be closed if all transitions preserve weight ($|s| = |s'|$).

2. Monotonicity and Well-Quasi-Ordering

Monotonicity is central, encapsulated by the simulation lemma: transitions can be "lifted" along the wqo. If $s \to s'$ and $s \preceq t$, then there exists a $t'$ with $t \to t'$ and $s' \preceq t'$. The existence of the finite-basis property for wqo-ordered state spaces ensures that any upward-closed set is determined by finitely many minimal elements. This, together with monotonicity, yields decidability for reachability of upward-closed targets by confining analysis to finitely many minimal predecessors (Aspnes, 24 Dec 2025).

3. Canonical Examples and Model Unification

Three primary distributed stochastic models fit the SWSTS paradigm:

• Population Protocols ($\POP$): States are multisets of agent states, transitions are random pairwise interactions, state ordering by subsequence embedding, with $\Pr(s \to s') = \Theta(n^{-2})$ for size $n$.
• Chemical Reaction Networks ($\CRN$): States are $\mathbb{N}^Q$ species count vectors, transitions correspond to reactions, componentwise ordering, rates conform to mass-action kinetics and are polynomial in configuration size.
• Synchronous Gossip/Token-Passing Models: Includes pull, push, matching, shuffle, and various betweenness-based variants; each adopts an appropriate wqo-compatible ordering and yields transition probabilities polynomially bounded below in system size.

These models (and their ordered or equivalence-augmented counterparts) all instantiate closed SWSTS under suitable scheduler assumptions, supporting the unification claim (Aspnes, 24 Dec 2025).

4. Temporal Dynamics: Phase Clocks and Computation Termination

For explicit signaling (as in phase-clock construction), any protocol in a closed SWSTS that attempts to alternate between two upward-closed targets is subject to the following: after a polynomial number of expected steps, either the system loses the ability to revisit the target (permanently fails) or else transitions between the sets occur in expected constant time (i.e., the clock "ticks too fast"). This is formalized in the "Finish-or-fail" theorem:

Let $\tau = \min\{i \mid X_i \in T \lor X_i \notin \Pre^*(T)\}$. For every initial configuration $X_0$, $\mathbb{E}[\tau] = O(|X_0|^k)$ for some fixed $k$. Any explicitly terminating protocol reaches a decision or enters a failure region in expected polynomial time. Contrariwise, stably convergent protocols need not terminate in polynomial expected time—convergence to a stable configuration may require exponential steps (Aspnes, 24 Dec 2025).

5. Computational Expressiveness and Simulations

SWSTS admit precise complexity-theoretic characterizations:

• Uniform SWSTS protocols (for which configurations and transitions are bitstring-encoded and sampleable in polynomial time) decide exactly the languages in BPP. That is, any uniform, closed SWSTS protocol with bounded error $\varepsilon < 1/2$ for a Boolean input language $L$ computes $L \in \mathrm{BPP}$.
• Symmetric, log-space-uniform SWSTS protocols decide exactly the symmetric languages in BPL. Symmetry entails indistinguishability of agents up to permutation and feasibility of encoding and simulation operations in logarithmic space.
• Conversely, every probabilistic Turing machine (PTM), hence every BPP computation, can be simulated in these models when augmented by a total order or equivalence relation via an oracle. Specifically,

$\mathrm{BPP} = \POP^< = \POP^= = \GOSSIP^< = \GOSSIP^= = \PUSH^< = \PUSH^= = \SHUFFLE^< = \SHUFFLE^= = \MATCHING^< = \MATCHING^= = \BETWEEN.$

The direct correspondence is achieved, for example, in ordered population protocols by encoding a single-leader, head-in-tape Turing machine simulation, and in unordered settings by simulating tape indices using equivalence classes (Aspnes, 24 Dec 2025).

6. Unified Theoretical Implications

The SWSTS construction bridges distributed stochastic computation and traditional complexity theory. Its principal abstraction—the stochastic scheduler with inverse-polynomial guarantees—enforces predictable computational and temporal properties for protocol termination and explicit signaling.

Key results include:

• Termination Bound: Any explicitly terminating SWSTS protocol halts in expected polynomial time.
• Computational Tightness: Uniform SWSTS exactly captures BPP; symmetric, log-space-uniform SWSTS capture symmetric BPL.
• Unified Modelling: Numerous models—population protocols, reaction networks, gossiping, token-matching, and their oracular extensions—are subsumed under the SWSTS framework (Aspnes, 24 Dec 2025).

These properties delineate a robust interface between population protocol theory, distributed random-interaction computation, and classical space/complexity classes, yielding both negative results (such as the impossibility of robust, evenly-paced long-running phase clocks) and positive ones (sharp computational expressiveness characterizations).