Alternating Weak Automata from Universal Trees

An improved translation from alternating parity automata on infinite words to alternating weak automata is given. The blow-up of the number of states is related to the size of the smallest universal ordered trees and hence it is quasi-polynomial, and only polynomial if the asymptotic number of priorities is logarithmic in the number of states. This is an exponential improvement on the translation of Kupferman and Vardi (2001) and a quasi-polynomial improvement on the translation of Boker and Lehtinen (2018). Any slightly better such translation would (if—like all presently known such translations—it is efficiently constructive) lead to algorithms for solving parity games that are asymptotically faster in the worst case than the current state of the art (Calude, Jain, Khoussainov, Li, and Stephan, 2017; Jurdzi'nski and Lazi'c, 2017; and Fearnley, Jain, Schewe, Stephan, and Wojtczak, 2017), and hence it would yield a significant breakthrough.

💡 Research Summary

The paper presents a new, more efficient translation from alternating parity automata (APA) on infinite words to alternating weak automata (AWA). The motivation stems from earlier constructions: the classic Kupferman‑Vardi translation incurs an exponential blow‑up of O(n·d) (n = number of states, d = number of priorities), while the more recent Boker‑Lehtinen approach reduces this to a quasi‑polynomial blow‑up of O(n·log n·log(d/ log n)). Both are far from optimal, especially when the number of priorities grows.

The authors’ key insight is to relate the state‑space blow‑up directly to the size of the smallest universal ordered tree (UOT) that can embed all (n, h)‑height ordered trees. Universal trees have become central in recent breakthroughs on parity‑game complexity (Calude‑Jain‑et al., Jurdziński‑Lazić). By constructing a translation whose blow‑up equals the size of an (n,⌊d/2⌋)‑universal tree, they achieve a bound that is polynomial in n when d = O(log n) and otherwise n·log(d/ log n)+O(1). This matches the best known quasi‑polynomial upper bounds for parity‑game solving.

Technically, the construction proceeds in two stages. First, the authors introduce a hierarchical decomposition of runs of an APA. Each run is broken into layers according to priority levels, and a “lazy progress measure” is assigned to vertices. This measure is a variant of the classic co‑Büchi progress measure: it allows an unbounded but finite number of odd‑priority repetitions, which can be recognized by a Büchi automaton. The lazy measure thus captures the essential parity condition while remaining amenable to Büchi acceptance.

Second, the lazy progress measures are mapped onto nodes of a universal tree. Because a universal tree contains every possible ordering of measure values up to the required height, any run of the original APA can be simulated by a run in a Büchi automaton whose state space corresponds to the tree’s nodes. The resulting Büchi automaton has exactly two priorities (a parity automaton with priorities {1,2}), i.e., it is a weak automaton after a final quadratic conversion.

The final step uses the well‑known quadratic conversion from alternating Büchi to alternating weak automata due to Kupferman‑Vardi. Since the intermediate Büchi automaton already has only a quasi‑polynomial number of states, the overall APA→AWA translation incurs a blow‑up of O(log(d/ log n)), a dramatic improvement over the previous O(log n·log(d/ log n)) factor.

The paper provides a fully constructive proof, avoiding the hybrid “parity‑weak” automata introduced in earlier work. The hierarchical decomposition is presented explicitly, and the universal‑tree embedding is shown to be sound and complete with respect to acceptance. Moreover, the authors discuss lower bounds: the best known Ω(n·log n) lower bound for APA→AWA translations (derived from Büchi complementation) is close to their upper bound, indicating near‑optimality. Consequently, any further asymptotic improvement would imply a breakthrough in parity‑game algorithms, surpassing the current quasi‑polynomial algorithms of Calude‑Jain‑et al., Jurdziński‑Lazić, and others.

In addition to the main translation, the paper revisits the co‑Büchi progress‑measure technique of Kupferman‑Vardi, clarifies its relationship to Klarlund’s progress measures, and introduces the lazy variant to handle finite but unbounded odd‑priority repetitions. The authors also acknowledge earlier work by Klarlund, noting that their proof is self‑contained and more transparent.

The conclusion outlines future directions: refining universal‑tree constructions to reduce constant factors, extending lazy progress measures to other acceptance conditions (e.g., Streett, Rabin), and performing empirical evaluations of the translation’s practical impact on model‑checking tools. Overall, the work significantly narrows the gap between the known upper and lower bounds for APA→AWA translations and deepens the connection between automata theory and parity‑game complexity.