A software testing approach to virtual reality applications using metamorphic testing

Author(s)

Stevão Alves de Andrade

• Document type: Doctoral Thesis
• Press: São Carlos.
• Institution: Universidade de São Paulo (USP). Instituto de Ciências Matemáticas e de Computação (ICMC/SB)
• Defense date: 2022-11-24

Examining board members

Márcio Eduardo Delamaro; José Remo Ferreira Brega; Cléber Gimenez Corrêa; Vinicius Humberto Serapilha Durelli

• Advisor: Márcio Eduardo Delamaro; Fátima de Lourdes dos Santos Nunes Marques

Abstract

Software Testing is one of the existing research areas in Software Engineering, being the main activity used to verify the conformity between software requirements and their respective implementations. The process of automating the software testing activity is a fundamental task that aims to give productivity and effectiveness. The automation of the software testing activity allows it to be conducted under systematic criteria, which guarantees the activitys reproducibility, in addition to significantly increasing the chance of identifying flaws in the evaluated product. Test oracles play a fundamental role within the testing activity, being responsible for evaluating the behavior of the outputs produced by a software during the testing activity. In this context, test oracles can be derived from software specifications, formal methods, assertions, machine learning techniques, metamorphic relationships, among others. This doctoral thesis proposes and evaluates a new automated software testing approach for Virtual Reality applications. Unlike conventional programs, virtual reality applications suffer from the test oracle problem, which occurs in situations where the outputs of the system under test are given in unconventional formats, such as machine learning models, images, three-dimensional objects and virtual reality environments. The purpose of the approach presented in this thesis is to use the metamorphic testing technique to create test constraints for virtual reality applications and to use reinforcement learning to enable the automatic generation of test data to automate the testing process of virtual reality applications. To do so, the work was conducted in three different stages: (i) the first stage consisted of a survey of software artifacts, available in open source repositories, in order to understand perceptions and extract models of practices of software quality in the context of virtual reality programs; in the second stage (ii) was a survey designed to investigate the perception of groups of interest, aiming to discover what is lacking about software quality practices in the context of virtual reality in order to clearly map the existing needs; and finally the third stage (iii) consisted in the development and evaluation of the testing approach based in the results observed in the previous steps. (AU)

• FAPESP's process: 17/19492-1 - Mutation testing applied to virtual reality environments modeled with scene graph
• Grantee: Stevão Alves de Andrade
• Support Opportunities: Scholarships in Brazil - Doctorate