Banca de DEFESA: MARIANNE BATISTA DINIZ DA SILVA

Uma banca de DEFESA de DOUTORADO foi cadastrada pelo programa.
STUDENT : MARIANNE BATISTA DINIZ DA SILVA
DATE: 26/05/2022
TIME: 14:00
LOCAL: Remoto
TITLE:

A data stream-driven methodology for driver behavior modeling.

KEY WORDS:

Internet of Things, Data Stream, Intelligent Vehicle, Driver Behavior, Online Unsupervised Learning.

PAGES: 100
BIG AREA: Engenharias
AREA: Engenharia Elétrica
SUBÁREA: Eletrônica Industrial, Sistemas e Controles Eletrônicos
SPECIALTY: Automação Eletrônica de Processos Elétricos e Industriais
SUMMARY:

The Internet of Things (IoT) is a growing network of objects — sensors, devices, systems, and more — that capture, transfer data, and communicate with each other through communication protocols. In addition, such objects have the ability to produce potentially unlimited sequences of data, called data streams. In this sense, it is clear that the IoT is creating new opportunities for various sectors, as can be seen in an unconventional way in the automotive sector. It is known that as a consequence of the advancement of architecture, vehicles are becoming increasingly equipped with various sensors and computational power. And, from available interfaces, it becomes possible to capture and extract, in an automated way, information through sensors and communication protocols present in vehicles and enabling a scenario known as the Internet of Intelligent Vehicles (IoIV). One of the benefits of IoIV is the creation of diagnostic applications, such as characterizing the behavior of drivers. This type of diagnosis is an essential requirement since the way you drive can impact different contexts, such as traffic safety, fuel consumption, emissions, and maintenance, among others. Furthermore, solutions generally available in the literature for analyzing drivers’ behavior focus on supervised offline learning models, fed with the entire dataset for training and testing. On the other hand, such solutions do not handle data streams suitable for online learning, that is, without knowledge of subsequent data. In face of this reality, the objective of this work is to identify patterns in the behavior of drivers, from a methodology oriented to data streams and unsupervised online algorithms. The methodology is adaptable and flexible, and considers the historical-temporal relationship between the samples, adapting in an autonomous and evolutionary way, without the need for a supervised training phase. In order to validate the proposed methodology, a case study was carried out in a real scenario with different conditions, which allowed the identification of daily driving operations. The results indicated the feasibility of the proposal regarding the identification of event detection and indicators of driver behavior. Therefore, the methodology can contribute to several applications, such as industry 4.0 — customized maintenance, fault detection — smart cities and urban mobility — improvement of pavements, increase in the number of speed reducers and crosswalks, decrease
in the maximum speed of roads, among others.

BANKING MEMBERS:
Presidente - 2885532 - IVANOVITCH MEDEIROS DANTAS DA SILVA
Interno - 1153006 - LUIZ AFFONSO HENDERSON GUEDES DE OLIVEIRA
Externo à Instituição - MAX MAURO DIAS SANTOS - UTFPR
Externo à Instituição - EDUARDO ALMEIDA SOARES
Externo à Instituição - JUAN MOISES MAURICIO VILLANUEVA - UFPB