System support for self-adaptive Cyber-Physical Systems

• Ano de defesa: 2015
• Autor(a) principal: Maia, Marcio Espíndola Freire
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Palavras-chave em Português: Internet of things; Middleware; Framework; Self-adaptation
• Link de acesso: http://www.repositorio.ufc.br/handle/riufc/56577

Devices with computational and communication capabilities are pervaded throughout the physical environment, generating information about users and their interactions to enrich the virtual world. This proximity between physical and cyber worlds has allowed the appearance of Cyber-Physical Systems (CPS), formed by users and devices (physical) interacting with services, components and applications (cyber). When the number of physical and cyber entities increase, so does the need for mechanisms to specify and manage their interactions. In that direction, this doctoral thesis proposes CyberSupport, an adaptable software stack to offer system support for selfadaptive CPS. The thesis contribution is based on layers: on the bottom, the communication and coordination layer is responsible for a modular, uncoupled and adaptable infrastructure to permit access and control of environment resources, along with message exchange and interaction of decentralized devices. On the top layer, the execution and adaptation layer permits to specify and implement monitoring and execution functionalities to support the creation of self-adaptive CPS. CyberSupport is evaluated according to performance metrics using real devices, along with design metrics using existing applications before and after using CyberSupport. The performance metrics evaluate 5 routing algorithms and 5 communication technologies implemented using CyberSupport according to message delivery and message loss metrics. Additionally, the design evaluation reimplementes 3 existing applications and compares the implementations with and without CyberSupport according to coupling metrics. The results shows an improvement of around 30% considering the design metrics.