Quantized State Hybrid Automata for Cyber-Physical Systems

TL;DR

This paper introduces Quantized State Hybrid Automata (QSHA), a novel formal model for simulating cyber-physical systems that ensures accurate detection of discontinuities with fewer simulation steps.

Contribution

The paper proposes a new dynamic quanta-based formal model, QSHA, improving accuracy and efficiency in simulating hybrid automata for cyber-physical systems.

Findings

QSHA guarantees accurate detection of level crossings.

QSHA reduces simulation steps significantly compared to existing methods.

Benchmarking shows improved efficiency in simulation.

Abstract

Cyber-physical systems involve a network of discrete controllers that control physical processes. Examples range from autonomous cars to implantable medical devices, which are highly safety critical. Hybrid Automata (HA) based formal approach is gaining momentum for the specification and validation of CPS. HA combines the model of the plant along with its discrete controller resulting in a piece-wise continuous system with discontinuities. Accurate detection of these discontinuities, using appropriate level crossing detectors, is a key challenge to simulation of CPS based on HA. Existing techniques employ time discrete numerical integration with bracketing for level crossing detection. These techniques involve back-tracking and are highly non-deterministic and hence error prone. As level crossings happen based on the values of continuous variables, Quantized State System (QSS)- integration may be more suitable. Existing QSS integrators, based on fixed quanta, are also unsuitable for simulating HAs. This is since the quantum selected is not dependent on the HA guard conditions, which are the main cause of discontinuities. Considering this, we propose a new dynamic quanta based formal model called Quantized State Hybrid Automata (QSHA). The developed formal model and the associated simulation framework guarantees that (1) all level crossings are accurately detected and (2) the time of the level crossing is also accurate within floating point error bounds. Interestingly, benchmarking results reveal that the proposed simulation technique takes 720, 1.33 and 4.41 times fewer simulation steps compared to standard Quantized State System (QSS)-1, Runge-Kutta (RK)-45, and Differential Algebraic System Solver (DASSL) integration based techniques respectively.