Kyveli Doveri,Pierre Ganty,B. Srivathsan
Abstract:The well-known Nerode equivalence for finite words plays a fundamental role in our understanding of the class of regular languages. The equivalence leads to the Myhill-Nerode theorem and a canonical automaton, which in turn, is the basis of several automata learning algorithms. A Nerode-like equivalence has been studied for various classes of timed languages.
In this work, we focus on timed automata with integer resets. This class is known to have good automata-theoretic properties and is also useful for practical modeling. Our main contribution is a Nerode-style equivalence for this class that depends on a constant K. We show that the equivalence leads to a Myhill-Nerode theorem and a canonical one-clock integer-reset timed automaton with maximum constant K. Based on the canonical form, we develop an Angluin-style active learning algorithm whose query complexity is polynomial in the size of the canonical form.
What problem does this paper attempt to address?
The core problem that this paper attempts to solve is to provide a characterization similar to the Myhill - Nerode theorem for Timed Automata with Integer Resets (IRTA), and construct an automaton in a canonical form based on this characterization. Specifically, the authors hope to find a Nerode - style equivalence relation depending on the constant \(K\) such that this equivalence relation can lead to the Myhill - Nerode theorem and can derive a 1 - clock integer - reset timed automaton with the maximum constant \(K\). Moreover, based on this canonical form, they also develop an Angluin - style active learning algorithm with polynomial query complexity.
### Background and Motivation of the Paper
In the process of understanding regular languages, the Myhill - Nerode theorem plays a crucial role. It classifies regular languages through the Nerode equivalence relation and provides the basis for minimizing deterministic finite automata. However, for temporal languages and timed automata, directly applying the Nerode equivalence relation will encounter challenges, mainly in the following aspects:
1. **Infinitely Many Equivalence Classes**: Temporal languages may have infinitely many quotient languages.
2. **States Cannot Be Uniquely Determined**: Two words with the same residual languages may never enter the same control state.
3. **Monotonicity Problem**: It is necessary to find a suitable definition of monotonicity to adapt to the construction of timed automata.
### Main Contributions
The main contribution of this paper is to propose a Nerode - style equivalence relation \(\approx_{L,K}\) applicable to timed automata with integer resets, and prove the following theorem:
\[ \text{The language } L \text{ is accepted by a timed automaton with integer resets if and only if there exists a constant } K \text{ such that } \approx_{L,K} \text{ is } K\text{-monotonic and has a finite index.} \]
This equivalence relation is coarser than any \(K\)-monotonic \(L\)-preserving equivalence relation. Intuitively, \(K\) represents the largest integer that may appear in the guard conditions in the accepting automaton.
### Canonical Form and Learning Algorithm
Based on the above characterization, the authors further derive the canonical form of IRTA and propose an algorithm that can calculate and learn this canonical form. In particular, they show how to adapt the Angluin - style learning method to learn 1 - IRDTA (1 - clock deterministic integer - reset timed automaton).
### Related Work
Existing research has explored the canonical forms of temporal languages, such as the machine - independent characterization proposed by Bojańczyk and Lasota [6], and the research on real - time languages by An et al. [3]. These studies either modify the automaton model or apply the characterization to specific classes of languages. This paper continues to use the classical timed automaton model and applies it to a new subclass - timed automata with integer resets.
### Conclusion
In general, this paper solves the characterization problem of timed automata with integer resets, provides a new equivalence relation and the corresponding canonical form, and provides theoretical basis and algorithm support for the learning of such automata.