Defesa de Tese de Doutorado – Sidney Roberto Dias de Carvalho – 30/9/2020

14/10/2020 18:11
Defesa de Tese de Doutorado
Aluno Sidney Roberto Dias de Carvalho
Orientador Prof. Ubirajara Franco Moreno, Dr. – DAS/UFSC
Data

 

30/9/2020  (quarta-feira) – 14h

Videoconferência

(https://meet.google.com/kbv-eozt-fzh)

 

 

Banca

Prof. Ubirajara Franco Moreno, Dr. – DAS/UFSC (presidente);

Prof. Jés de Jesus Fiais Cerqueira, Dr. – DEE/UFBA;

Prof. Eugênio de Bona Castelan Neto, Dr. – DAS/UFSC;

Prof. André Ricardo Fioravanti, Dr. – DMC/UNICAMP;

Título Network Topology Control for Connectivity Maintenance and Information Spreading Manipulation in Multi-Robot Systems
Abstract: Cooperation is a desired property for many interconnected systems since it allows them to solve complex tasks distributively. The increase in the use of cooperative processes (characterised by information sharing) decreases the individual effort demanded by each agent. Nevertheless, this decrease of local effort increases the system complexity as a whole due to the intrinsic couplings between the agents that could give rise to emerging behaviours, which are highly dependent on the communication topology. One of these behaviours, called consensus, is necessary to ensure cohesion among all the exchanged information. This property is an essential and desirable feature in multi-agent systems that are employed to execute cooperative tasks since their individuals must share pieces of information that ensure their teammates are trying to solve the same global problem. Thus, to keep the cooperation, an interconnected system must ensure cohesion through consensus-based diffusion dynamics. In such a context, this work aims two distinct objectives by manipulating the network topology of an interconnected system composed of autonomous agents: i) ensure the system connectivity even when there are faults in one agent, during the performed cooperative task. ii) control the information spreading to drives the consensus convergence, according to the intended information distribution of the system. This thesis undertakes both objectives separately. It solves the first objective through the solution of the Travelling Salesman Problem (TSP) applied over indicators of signal strength in wireless networks to create virtual bi-connected topology in a multi-agent system. Then, it uses a Model Predictive Controller (MPC) executed in a decentralised way to move the agents toward each other, turning the virtual links into real ones and bi-connecting the network topology of the system. This procedure can turn any connected network into a fault-tolerant one. During the procedure, an approach based on Sequential Convex Programming (SCP) applied over the aforementioned MPC framework ensures non-collision among the agents. For the second objective, this thesis uses an approach based on Semidefinite Programming (SDP) to design the optimal weights of a network adjacency matrix, in order to control the convergence of a distributed random consensus protocol for variables at the discrete-space domain. It uses Markov theory and the biological inspiration of epidemics to find out a dynamical spreading model that can predict the information diffusion under this consensus protocol. Also, it presents convergence properties and equilibrium points of the proposed model regarding the network topology. Finally, extensive numerical simulations and experiments performed in a commercial robotic platform evaluate the effectiveness of the proposed approaches for both objectives.