Decentralized Predicate Detection over Partially Synchronous Continuous-Time Signals

TL;DR

This paper introduces the first decentralized algorithm for detecting safety predicate violations in continuous-time signals within partially synchronous cyber-physical systems, addressing challenges of partial knowledge, clock drift, and continuous signals.

Contribution

It presents a novel decentralized detection algorithm for conjunctive predicates over continuous-time signals in partially synchronous systems, with proven complexity and experimental validation.

Findings

Algorithm detects all predicate violations including clock drift effects.

Detection complexity matches that of discrete systems.

Experimental validation confirms practical effectiveness.

Abstract

We present the first decentralized algorithm for detecting predicates over continuous-time signals under partial synchrony. A distributed cyber-physical system (CPS) consists of a network of agents, each of which measures (or computes) a continuous-time signal. Examples include distributed industrial controllers connected over wireless networks and connected vehicles in traffic. The safety requirements of such CPS, expressed as logical predicates, must be monitored at runtime. This monitoring faces three challenges: first, every agent only knows its own signal, whereas the safety requirement is global and carries over multiple signals. Second, the agents' local clocks drift from each other, so they do not even agree on the time. Thus, it is not clear which signal values are actually synchronous to evaluate the safety predicate. Third, CPS signals are continuous-time so there are potentially uncountably many safety violations to be reported. In this paper, we present the first decentralized algorithm for detecting conjunctive predicates in this setup. Our algorithm returns all possible violations of the predicate, which is important for eliminating bugs from distributed systems regardless of actual clock drift. We prove that this detection algorithm is in the same complexity class as the detector for discrete systems. We implement our detector and validate it experimentally.