On Reconfigurable Bisimulation, with an Application to the Distributed Synthesis Problem

2025-05-27

AI Summary

Centralized migration systems face challenges in transitioning to asynchronous, agent-based distributed implementations due to rigid communication constraints requiring predefined event-to-agent mappings. Method: We propose a parameterized reconfigurable bisimulation theory that enables on-demand dynamic adjustment of event participation, retaining only essential inter-agent communication. Integrated with state-space minimization and dynamic communication pruning, our approach preserves behavioral equivalence while substantially reducing the number of distributed agents. Contribution/Results: This work achieves, for the first time in distributed synthesis, a decidable solution to global specification enforcement under asynchronous distributed control. Experimental case studies demonstrate a significant expansion of the solvability frontier—i.e., a measurable increase in the class of specifications provably realizable by distributed agents.

Technology Category

• Multiagent Systems: Distributed Problem Solving
• Search and Optimization: Distributed Search
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

We consider the problem of distributing a centralised transition system to a set of asynchronous agents recognising the same language. Existing solutions are either manual or involve a huge explosion in the number of states from the centralised system. The difficulty arises from the need to keep a rigid communication scheme, specifying a fixed mapping from events to those who can participate in them. Thus, individual agents need to memorise seen events and their order to dynamically compare their knowledge with others when communicating. To bypass this, we rely on reconfigurable communication: agents decide locally ``by-need'' when to participate or discard specific events during execution while not impacting the progress of the joint computation. Our distribution relies on a novel notion of Parametric Reconfigurable Bisimulation, that identifies the only required participations. We show how to compute this bisimulation and that such minimisation produces a joint system that is bisimilar to the original centralised one. We use a case study to show its effectiveness by producing agents that are much smaller than the centralised system and jointly perform the same computations. As a notable application, we use this distribution in order to allow for distributed synthesis from global specifications. In this case, rigid communication leads to undecidability, which is bypassed by our ability to dynamically prune communications.

Problem

• Distribute centralized transition system to asynchronous agents efficiently
• Minimize state explosion while maintaining language recognition
• Enable dynamic communication pruning for distributed synthesis

Innovation

• Reconfigurable communication for dynamic participation
• Parametric Reconfigurable Bisimulation for minimal participations
• Distributed synthesis with dynamically pruned communications