Defesa de Exame de Qualificação – Sidney Roberto Dias de Carvalho – 22/08/2017
22/08/2017 17:02
| Defesa de Exame de Qualificação | |
| Aluno | Sidney Roberto Dias de Carvalho |
| Orientador | Prof. Ubirajara Franco Moreno, Dr. – DAS/UFSC |
| Data
Local |
22/08/2017 09h00 (terça-feira)
Sala PPGEAS II (piso inferior) |
| Prof. Eugênio de Bona Castelan Neto, Dr. – DAS/UFSC (presidente)
Prof. André Gustavo Scolari Conceição, Dr. – UFBA Prof. Adão Boava, Dr. – ENG/UFSC |
|
| Título
|
Controle de Topologia Aplicada a Problemas de Consenso |
| Resumo: Cooperative tasks demand less individual effort to be solved when compared to regular ones. However, this decrease of local effort increases the communication complexity. The information sharing process is an essential and desirable feature in multi-robot systems that are employed to execute cooperative tasks. In such systems, the individuals must share information with some reliability to ensure their teammates are trying to solve parts of the same global problem. So, information cohesion is an essential key to the correct actuation of interconnected systems under cooperative strategies. Such property can be ensured when the communication reaches some configurations obtained by correctly managing the network topology. In this work, we want to manipulate the network topology of a multi-robot system composed of autonomous mobile robots in a way that the resulting topology will be fault tolerant to the dropout of other robots, and to the eventual exogenous disturbances that can happen. The considered system is fully decentralised and distributed in the sense that there is no base-station or leader in it, and the known information is local, restricted to each robot’s neighbourhood. We use algebraic graph properties and consensus theory to develop an approach that can turn any connected network into a fault tolerant one. Besides that, some studies are performed to evaluate the relationship between the epidemic description of information spreading, the frequency response of interconnected systems under consensus dynamics and the fundamental network components, such as vertices and edges. Using this information, we present some strategies to solve the problem of disturbance attenuation in consensus-based networks in a distributed way while the connectivity is ensured. | |
