A process for sound conformance testing of cyber-physical systems

• Autor(a) principal: ARAUJO, Hugo Leonardo da Silva
• Data de Publicação: 2017
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Institucional da UFPE
• Texto Completo: https://repositorio.ufpe.br/handle/123456789/29687
• topic: Engenharia de software; Métodos formais

The term Hybrid System is used to describe a modelling formalism of systems that combine discrete and continuous aspects; for instance, a system where a controller (discrete component) is connected to a physical system (continuous component). Systems that encompass tightly integrated digital and physical components and deal with spatial and temporal metrics, besides involving human interaction, are known as Cyber-Physical Systems (CPS). Model-based testing of CPSs is a recent subject in the literature, and it is still being actively researched and developed. The analysis of CPSs is usually complex due to their multidisciplinary nature, with such systems dealing with aspects of different subject areas such as computer science, physics and control systems. In this work, we propose a process for sound conformance testing of cyberphysical systems. The main goal of this process is to provide a practical and semi-automatic solution to testing CPSs. Some of the steps of our process were mechanized through the use of a prototype tool that we have developed. This project was conceived during the literature review in our research when we realized the absence of a structured process with systematic steps for conformance testing of CPSs. We first focused on studying the existing conformance testing strategies of hybrid systems and settled on working with (τ, ε)-conformance relation. In this conformance notion, the outputs of both specification and implementation models are compared under the same input stimuli. It makes use of temporal (τ) and spatial (ε) margins of error to determine if the output behaviours are close enough to each other. In conformance verification strategies based on this relation, an issue related to soundness was brought to our attention, which made us shift our focus to solve this problem through reachability analysis. We noticed that the sampling rate, used to observe the system behaviour at discrete points, was closely related to the soundness problem identified. This motivated the definition and partial automation of a process to support conformance testing of CPSs. The proposed process involves five steps: (i) automatic sampling rate computation; (ii) margins of error definition (temporal and spatial); (iii) performing reachability analysis to obtain sound verdicts; (iv) conformance testing (test generation, test execution and verdict attainment); (v) result analysis and parameters tuning. Additionally, we have performed an empirical analysis to shown how our approach can be used in practice describing a few usage scenarios as well as implementing two case studies: a combustion engine controller and a pneumatic suspension system.