Automatic Generation of Metamorphic Relations for a Cyber-Physical System-of-Systems Using Genetic Algorithm

Abstract

A Cyber-Physical System-of-Systems (CPSoS) has innate uncertainties from operation in the physical environment and interaction among the constituent systems. These uncertainties make a CPSoS more susceptible to the oracle problem, a challenge in determining the correct behavior when testing the system. Metamorphic testing (MT) suggests a solution to addressing this challenge by utilizing metamorphic relations (MRs), relations among multiple inputs and corresponding outputs of the system. However, when applying MT on a CPSoS, generating MRs is difficult due to the continuous operation of a CPSoS in uncertain environment. In this study, we propose a method to automatically generate MRs from field operational test (FOT) data logs of a CPSoS. We define an MR template to capture the CPSoS behaviors. We then apply genetic algorithm to adapt the MR generated by the engineers, and thus improve the testing effectiveness. Our method is validated in a case study of an autonomous robot vehicle. Our results show that the automatically generated MRs capture the behaviors of a CPSoS more realistically than the manually generated MRs. With our method, engineers can obtain CPSoS MRs with minimal manual effort.