Intelligent decision-making solutions for residential infrastructures

Author(s)

Geraldo Pereira Rocha Filho

• Document type: Doctoral Thesis
• Press: São Carlos.
• Institution: Universidade de São Paulo (USP). Instituto de Ciências Matemáticas e de Computação (ICMC/SB)
• Defense date: 2018-05-18

Examining board members

Jo Ueyama; Daniel Macedo Batista; Kelvin Lopes Dias; Rudinei Goularte

• Advisor: Jo Ueyama

Abstract

In recent years, energy efficiency has become a major global challenge, and energy waste is a factor that needs to be highlighted. Such waste can be overcome with the use of Home Automation System (HAS). It should be stressed that the HASs are strongly dependent on its internal network, since this is the basis of a smart home. Wireless Sensor and Actuator Networks (WSANs) provide a modern and ubiquitous infrastructure for a smart home. However, the use of WSANs to monitor and act (i.e. decision-making process) as a control infrastructure within the context of HAS poses a new problem. Such problem refers not only to the lack of a method to execute the decision-making process within the WSAN, but also to the lack of investigating a trade-off between the decision-making accuracy and the extension of the WSAN nodes life-time. In addition, the lack of a distributed infrastructure, with low overhead in processing and that reduces service latency are some of the new problems to be addressed in the literature. With this, one has as a challenge to embark on greater intelligence in devices with scarce resources, a feature present in a WSAN. To overcome such limitations, this thesis presents two intelligent decision-making solutions for residential infrastructures, named ResiDI and ImPeRIum. ResiDI was developed based on a neural network to act in the decision-making process within the network, as well as a temporal correlation mechanism to maximize the energy consumption in the networks nodes. ImPeRIum was based on a heterogeneous set of smart objects to form a fog computational environment, which manages the applications of the residence through a neural network. The solutions were evaluated extensively in different scenarios and compared with an approach in the literature. The real and simulated results, evaluated through parametric and non-parametric tests, show that solutions make four key contributions: (i) increased decisionmaking; (ii) reduction in node energy consumption; (iii) reduction in action response time with low overload; and (iv) efficiency in the transmission of information. (AU)

• FAPESP's process: 14/06330-5 - An approach of intelligent decision for an infrastructure distributed home automation using wireless sensors and actuators
• Grantee: Geraldo Pereira Rocha Filho
• Support Opportunities: Scholarships in Brazil - Doctorate