Deep Learning in Automated Tests for the Automotive Industry

• Autor(a) principal: Branco, Beatriz Cunha
• Data de Publicação: 2023
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10400.22/23878
• topic: Artificial Intelligence; Machine Learning; Deep Learning; Automotive Sector; Automated Software Testing; Android Automotive; Python; Inteligência Artificial; Setor Automóvel; Testes de Software Automatizados

Artificial Intelligence (AI) usage has increased over the years, whether in speech recognition, predictive analytics and mainly image recognition, being useful for industries such as the medical industry, for medical diagnosis, the automotive industry, for autonomous driving, traffic signals detection and also for an advanced driver assistance systems that can help prevent accidents. The automotive industry is one of the most popular industries in our society, either by the convenience that it gives to people on their travels, the designs of the vehicles themselves and nowadays, mostly the technology that it offers. And, with this increase in technology the rise of technical failures also occurs, which makes one step of the production of any kind of vehicle more important than ever, testing, which caused the companies to start investing in automated tests in order to decrease both their costs in a long term and human effort. There are a lot of techniques when it comes to automated testing, but with the increase in the popularity of Deep Learning (DL) and Machine Learning (ML), automated software testing using this branch of AI has started to gain popularity as well. Therefore, the project documented in this report has the goal of using a DL network that can help to perform automated software tests in Android Automotive infotainment together with Python and its dedicated frameworks, which can then increase the quality of testing but also decrease the human effort. Besides this, reports with the results of the executions and the suggestion of tests’ names based on the network output will be implemented.