Uma abordagem para validação e recuperação de erros em dados de sensores em aplicações para cidades inteligentes

Abstract

Smart Cities connect their physical, technological, social, and business infrastructures to improve citizens’ well-being. Cities have experienced significant growth over the years, driving the need for Smart City Applications. Due to the heterogeneous environment of such applications, failure scenarios may never be tested. Therefore, it is important to employ efficient fault-tolerance techniques that ensure the reliability of the data used by these applications. Objective: To model and evaluate an approach for validating and recovering errors in sensor data from Smart City Applications, focusing primarily on the data validation stage, that is, on detecting incorrect data from sensors in Smart City Applications. Methodology: To achieve the general objective, the following steps were carried out: (i) an analysis of the state of the art regarding fault-tolerance techniques in Smart City Applications; (ii) the design of an approach for error validation and recovery suitable for Smart City scenarios; (iii) the selection and implementation of algorithms for sensor data validation; and (iv) the evaluation of the algorithms Isolation Forest, Support Vector Machines (SVM) One-Class, and Correlation-Based Diversity proposed for error detection. Results and Conclusions: As theoretical contributions, the state of the art on fault-tolerance techniques within the investigated context was characterized, and an approach was conceived to validate and correct data originating from sensors in Smart City Applications. As practical contributions, software functionalities were implemented to enable the execution of experimental studies on error detection in sensor data. Regarding the experiments conducted, all algorithms showed good performance in detecting errors for the evaluated application scenario, with Correlation-based Diversity achieving the best performance in the shortest execution time.