Abstract:Branching bisimilarity on normed BPA processes was recently shown to be decidable by Yuxi Fu (ICALP 2013) but his proof has not provided any upper complexity bound. We present a simpler approach based on relative prime decompositions that leads to a nondeterministic exponential-time algorithm; this is close to the known exponential-time lower bound.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is about the complexity of branching bisimulation equivalence in normalized BPA (Basic Process Algebra) processes. Specifically, Yuxi Fu proved at the ICALP conference in 2013 that the branching bisimulation equivalence of normalized BPA processes is decidable, but his proof did not provide any upper - bound complexity estimate. The authors of this paper propose a new method based on relatively prime decomposition, which leads to a non - deterministic exponential - time algorithm; this is close to the known exponential - time lower bound.
### Core problems of the paper
1. **Decidability and complexity of branching bisimulation equivalence**:
- Branching bisimilarity is an important concept of process equivalence, especially when dealing with systems with silent actions (τ - actions).
- A normalized BPA process refers to a process in which each variable can be gradually rewritten as the empty word.
- Yuxi Fu has proved that the branching bisimulation equivalence of normalized BPA processes is decidable, but has not given an upper bound on the complexity.
2. **Proposing a new algorithm and complexity analysis**:
- The authors of this paper propose a method based on relatively prime decomposition, simplify Fu's proof, and provide a specific upper bound on the complexity.
- Through this method, they show that the branching bisimulation equivalence of normalized BPA processes can be decided within non - deterministic exponential time (NExpTime).
### Formulas and definitions
- **Branching bisimulation equivalence**: Given a labeled transition system \( L=(S, A, (\xrightarrow{a})_{a\in A}) \), a symmetric relation \( B\subseteq S\times S \) is a branching bisimulation if for any \( (s, t)\in B \), \( a\in A \), and \( s'\in S \) such that \( s\xrightarrow{a}s' \), then:
- \( a = \tau \) and \( (s', t)\in B \), or
- \( a\in A_{vis}\cup\{\tau\} \) and there exists a sequence \( t = t_0\xrightarrow{\tau}t_1\xrightarrow{\tau}\cdots\xrightarrow{\tau}t_k\xrightarrow{a}t' \) (for some \( k\geq0 \)) such that \( (s', t')\in B \) and \( (s, t_i)\in B \) for all \( i\in\{1, 2,\ldots, k\} \).
- **Class - change norm**: The class - change length (cc - length) of a path \( \alpha w\xrightarrow{}\beta \) is the number of class - change transitions in the path. The class - change norm \( \langle\langle\alpha\rangle\rangle \) is the minimum cc - length of the path \( \alpha w\xrightarrow{}\epsilon \).
- **Relatively prime decomposition**: For any \( R\subseteq V \), the \( R \)-prime form \( PDR(\alpha) \) of a configuration \( \alpha \) is the unique \( R \)-prime form configuration such that \( \alpha\sim_R PDR(\alpha) \).
### Summary
The main contribution of this paper lies in simplifying the problem of deciding the branching bisimulation equivalence of normalized BPA processes by introducing the method of relatively prime decomposition and limiting its complexity within non - deterministic exponential time.