A simulation-based software environment for verifying control and decision-making algorithms in a multi-agent cyber-physical system

Abstract

The paper presents the architecture of a simulation-based software environment for verifying control and decision-making algorithms in multi-agent cyber-physical systems. The environment integrates models of interacting agents, communication channels, distributed data storage, and a controlled technological process, together with tools for scenario generation, software fault injection, event logging, and automated analysis of experimental results. Each agent is described by static and dynamic parameters representing its functional state, computational, energy, and communication resources, locally stored and network-accessible data, and assigned role within a collective task. The communication model accounts for delays, bandwidth limitations, message loss, corruption, and duplication. A formal set of nominal and stress-test scenarios is defined to reproduce agent hardware and resource failures, network degradation, changes in process parameters, and combined disruptive effects. Experimental outcomes are evaluated using indicators of fault tolerance, fault detection and recovery time, behavioural predictability, reproducibility, preservation and accessibility of physical data, timely and correct message delivery, and resource-utilisation balance. The proposed environment supports systematic assessment of the reliability, resilience, and efficiency of distributed control algorithms under diverse operating conditions.