Optimising Highly-Parallel Simulation-Based Verification of Cyber-Physical Systems

TL;DR

This paper introduces a parallel simulation optimization technique for cyber-physical systems verification, significantly reducing the time needed for large-scale scenario-based simulations in high-performance computing environments.

Contribution

It proposes a method to compute optimal parallel simulation campaigns that minimize total verification time by avoiding redundant simulations, enabling practical large-scale system verification.

Findings

Achieved up to 8x speedup in simulation campaigns.

Enabled verification of up to 200 million scenarios within weeks.

Demonstrated effectiveness on Modelica, FMU, and Simulink models.

Abstract

Cyber-Physical Systems (CPSs), comprising both software and physical components, arise in many industry-relevant domains and are often mission- or safety-critical. System-Level Verification (SLV) of CPSs aims at certifying that given (e.g., safety or liveness) specifications are met, or at estimating the value of some KPIs, when the system runs in its operational environment, i.e., in presence of inputs (from users or other systems) and/or of additional, uncontrolled disturbances. To enable SLV of complex systems from the early design phases, the currently most adopted approach envisions the simulation of a system model under the (time bounded) operational scenarios of interest. Simulation-based SLV can be computationally prohibitive (years of sequential simulation), since model simulation is computationally intensive and the set of scenarios of interest can huge. We present a technique that, given a collection of scenarios of interest (extracted from mass-storage databases or from symbolic structures, e.g., constraint-based scenario generators), computes parallel shortest simulation campaigns, which drive a possibly large number of system model simulators running in parallel in a HPC infrastructure through all (and only) those scenarios in the user-defined (possibly random) order, by wisely avoiding multiple simulations of repeated trajectories, thus minimising the overall completion time, compatibly with the available simulator memory capacity. Our experiments on Modelica/FMU and Simulink case study models with up to ~200 million scenarios show that our optimisation yields speedups as high as 8x. This, together with the enabled massive parallelisation, makes practically viable (a few weeks in a HPC infrastructure) verification tasks (both statistical and exhaustive, with respect to the given set of scenarios) which would otherwise take inconceivably long time.