bisimulation-checker

name: bisimulation-checker description: 'Checks bisimulation for process calculi. Use when: (1) Proving equivalence, (2) Compiler optimization, (3) Protocol verification.' version: 1.0.0 tags:

• semantics
• bisimulation
• equivalence
• process-calculus difficulty: intermediate languages:
• python
• ocaml
• coq dependencies:
• operational-semantics-definer

Bisimulation Checker

Checks bisimulation equivalence for concurrent and reactive systems.

When to Use

• Proving process equivalence
• Verifying compiler transformations
• Protocol verification
• Semantic equivalence

What This Skill Does

1. Defines bisimulation - Strong and weak
2. Checks equivalence - Via game characterization
3. Handles branching - With silent steps
4. Optimizes - Partition refinement

Core Theory

Strong Bisimulation:
  P ~ Q iff
    ∀a. P →a P' ⇒ ∃Q'. Q →a Q' and P' ~ Q'
    ∀a. Q →a Q' ⇒ ∃P'. P →a P' and P' ~ Q'

Weak Bisimulation:
  P ≈ Q iff
    ∀a. P =a= P' ⇒ ∃Q'. Q =a= Q' and P' ≈ Q'
  
  Where =a= is weak transition (including ε)

Implementation

from dataclasses import dataclass, field
from typing import Dict, Set, List, Optional, Tuple
from collections import defaultdict

@dataclass
class Process:
    """Process term"""
    pass

@dataclass
class Action:
    """Transition action"""
    label: str
    is_silent: bool = False  # τ/ε

@dataclass
class Transition:
    """Transition relation"""
    process: Process
    action: Action
    target: Process

class BisimulationChecker:
    """Check bisimulation"""
    
    def __init__(self):
        self.transitions: Dict[Process, List[Transition]] = defaultdict(list)
    
    def add_transition(self, p: Process, a: Action, q: Process):
        """Add transition p --a--> q"""
        
        self.transitions[p].append(Transition(p, a, q))
    
    def strong_bisimulation(self, p: Process, q: Process) -> bool:
        """
        Check strong bisimulation P ~ Q
        
        Algorithm: Partition refinement
        """
        
        # Initial partition: observable vs silent
        # Refine until stable
        # O((n+m) log n)
        
        return self._check_bisim(p, q, weak=False)
    
    def weak_bisimulation(self, p: Process, q: Process) -> bool:
        """
        Check weak bisimulation P ≈ Q
        """
        
        return self._check_bisim(p, q, weak=True)
    
    def _check_bisim(self, p: Process, q: Process, weak: bool) -> bool:
        """Generic bisimulation check"""
        
        # BFS with visited
        visited: Set[Tuple[Process, Process]] = set()
        worklist = [(p, q)]
        
        while worklist:
            r, s = worklist.pop()
            
            if (r, s) in visited:
                continue
            visited.add((r, s))
            
            # Check: r's transitions match s's
            r_trans = self.transitions.get(r, [])
            s_trans = self.transitions.get(s, [])
            
            for tr in r_trans:
                # Find matching transition in s
                if weak and tr.action.is_silent:
                    # Can match ε
                    worklist.append((tr.target, s))
                    continue
                
                found = False
                for st in s_trans:
                    if weak and st.action.is_silent:
                        found = True
                        worklist.append((r, st.target))
                        continue
                    
                    if tr.action.label == st.action.label:
                        found = True
                        worklist.append((tr.target, st.target))
                        break
                
                if not found:
                    return False
            
            # Check reverse
            for st in s_trans:
                if weak and st.action.is_silent:
                    worklist.append((r, st.target))
                    continue
                
                found = False
                for tr in r_trans:
                    if weak and tr.action.is_silent:
                        found = True
                        worklist.append((st.target, r))
                        continue
                    
                    if st.action.label == tr.action.label:
                        found = True
                        worklist.append((st.target, tr.target))
                        break
                
                if not found:
                    return False
        
        return True

# Process calculus syntax
class PCProcess(Process):
    """Process calculus processes"""
    
    def __init__(self, name: str):
        self.name = name

def pi_calculus_example():
    """
    Process definitions:
      P = x(y).P'    -- receive x then P'
      P = x̅⟨y⟩.P'   -- send y on x then P'
      P = τ.P'        -- silent action
      P | Q           -- parallel
      !P              -- replication
    
    Bisimulation checks:
      x(y).P | x̅⟨z⟩.Q ~ P{y/z} | Q
    """
    pass

Partition Refinement

class EfficientBisim:
    """O(n log n) bisimulation via partition refinement"""
    
    def bisim(self, processes: Set[Process]) -> List[Set[Process]]:
        """
        Partition processes into bisimulation equivalence classes
        """
        
        # Initial partition: by observable actions
        blocks = self._initial_partition(processes)
        
        # Refine until stable
        changed = True
        while changed:
            changed = False
            
            for block in blocks:
                # Try split
                splits = self._refine(block, blocks)
                if splits:
                    blocks = [b for b in blocks if b != block] + splits
                    changed = True
                    break
        
        return blocks
    
    def _initial_partition(self, procs: Set[Process]) -> List[Set[Process]]:
        """Initial partition by action profile"""
        
        profile = {}
        for p in procs:
            actions = frozenset(t.action.label for t in self.transitions[p])
            if actions not in profile:
                profile[actions] = set()
            profile[actions].add(p)
        
        return list(profile.values())
    
    def _refine(self, block: Set[Process], blocks: List[Set[Process]]) -> List[Set[Process]]:
        """Refine block based on predecessor blocks"""
        
        # For each action, check if targets differ
        # Split if necessary
        
        splits = []
        # ... implementation
        return splits

Applications

• Compiler optimization verification
• Protocol equivalence
• Refinement checking
• Process algebra

Key Concepts

Canonical References

Tips

• Use partition refinement
• Handle τ steps carefully
• Consider weak equivalence
• Optimize for large systems

Related Skills

• operational-semantics-definer - Semantics
• model-checker - Model checking
• session-type-checker - Session types

Research Tools & Artifacts

Real-world bisimulation and process verification tools:

Key Systems to Study

• CCS semantics: Formal basis for concurrency
• π-calculus: Mobile processes
• LOTOS: Formal description technique

Research Frontiers

Current active research in bisimulation:

Hot Topics

1. Bisimulation for security: Non-interference verification
2. Quantitative: Performance-aware equivalence
3. Distributed systems: Fault-tolerant equivalence

Implementation Pitfalls

Common bugs in bisimulation checkers:

The "τ vs ε" Bug

Weak bisimulation confuses silent steps:

• τ is internal action (must match)
• ε is empty trace (can be inserted)
• Correct: τ can simulate ε but not vice versa

Partition Refinement Complexity

O(n log n) via:

• Initial partition by observable profile
• Split blocks based on predecessor blocks
• Iterate until stable