Simulation-Based Metamorphic Testing for System Verification of Cyber-Physical Systems - A Systematic Assessment of Seamless EV Charging Systems Under Dynamic Conditions

Abstract

The increasing adoption of electric vehicles has driven demand for reliable public charging infrastructure, as well as the backend software systems that support it. Seamless charging services, which link charging sessions to vehicles by correlating asynchronous event streams from charge point operators and vehicle systems, introduce a verification challenge that conventional testing approaches are not well suited to address. There is no precomputable ground truth against which each matching decision can be evaluated, and reproducing relevant adversarial conditions through physical testing is not feasible at scale. This thesis presents the design, implementation, and evaluation of a simulation based metamorphic testing framework that addresses these constraints. The frame work combines discrete-event simulation, which generates realistic and adversarial event sequences in a reproducible execution environment, with metamorphic relations, which act as a substitute oracle by defining behavioral properties that must hold across related executions. Eight metamorphic relations that characterize correct vehicle-to-session matching behavior were derived through a four-step process consisting of source code analysis, brainstorming, peer review with the development team, and consolidation. The framework was evaluated against the Seamless Charging Service operated by WirelessCar AB across 18 scenarios. The simulator generated 467 validation reports. Six of the eight metamorphic relations held across the dataset, while two, matching exclusivity and match quality monotonicity, revealed 43 violations across 11 reports. These violations were traceable to structural inconsistencies in the system’s observed behavior, despite the system producing zero incorrect matches across 998 attempted matches. No confirmed false positives were observed across 397 control executions. The findings demonstrate that simulation-based metamorphic testing can detect meaningful behavioral deviations in event-driven cyber-physical systems without access to a traditional test oracle. At the same time, the strength of this conclusion is bounded by the conditions the framework was able to execute: the densest concurrency and event ordering scenarios did not complete within the configured execution budget, leaving the system’s behaviour under the highest-stress conditions unverified.