Probabilistic and Stateful Extensions to the π-Calculus for Formal Verification of AI-Driven Mobile and Communicating Systems
Journal of Algorithms in Complex Systems · 2026
DOI: 10.66879/jacs.2026.17 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Algorithms in Complex Systems, 2026. Verfasst von Chang Yang.
Abstract (Originalsprache)
AI-driven mobile and communicating systems—such as large language model (LLM) agents and autonomous drone swarms—present fundamental challenges for formal verification, arising from probabilistic behaviours, dynamic communication topologies, and continuous neural dynamics. Conventional static model checkers and component level neural verifiers frequently encounter severe scalability and expressiveness limitations when addressing these complex, state-augmented agent interactions. To bridge this gap, we introduce a unified probabilistic and stateful extension of the π-calculus, incorporating three foundational constructs: probabilistic action choice, finite sound abstraction of continuous neural dynamics, and bounded dynamic name generation for evolving topologies. We develop a strict meta-theory establishing that the behavioural distance db is a 1-bounded pseudometric, so that exact probabilistic bisimilarity (~0) is an equivalence relation, and that, for every ε ≥ 0, quantitative bisimilarity ~ε is a congruence with respect to every process operator other than replication; within the replication-free fragment of the calculus, exact bisimilarity (~0) is moreover decidable in polynomial time O (M log N) under finite-state quotient semantics. Operating orthogonally to component-level neural verifiers, the framework bypasses micro-level state-space explosion, enabling tractable protocol-level verification of dynamic multi-agent systems. This work demonstrates that formal abstraction constitutes a sound and scalable foundation for reasoning about the protocol-level safety and liveness of next-generation AI collectives.
Zitieren
APA
Yang, C. (2026). Probabilistic and Stateful Extensions to the π-Calculus for Formal Verification of AI-Driven Mobile and Communicating Systems. Journal of Algorithms in Complex Systems. https://doi.org/10.66879/jacs.2026.17
Vancouver
Yang C. Probabilistic and Stateful Extensions to the π-Calculus for Formal Verification of AI-Driven Mobile and Communicating Systems. Journal of Algorithms in Complex Systems. 2026. doi:10.66879/jacs.2026.17
BibTeX
@article{Yang2026,
author = {Yang, Chang},
title = {Probabilistic and Stateful Extensions to the π-Calculus for Formal Verification of AI-Driven Mobile and Communicating Systems},
journal = {Journal of Algorithms in Complex Systems},
year = {2026},
doi = {10.66879/jacs.2026.17}
}
EUVIMED