Controlling collective behaviors and autonomous mobile agents

• Ano de defesa: 2019
• Autor(a) principal: Vander Luis de Souza Freitas
• Orientador(a): Elbert Einstein Nehrer Macau, Serhiy Yanchuk
• Banca de defesa: Lamartine Nogueira Frutuoso Guimarães, Solon Venâncio de Carvalho, Patricia Romano Cirilo, José Roberto Castilho Piqueira
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Instituto Nacional de Pesquisas Espaciais (INPE)
• Programa de Pós-Graduação: Programa de Pós-Graduação do INPE em Computação Aplicada
• País: BR
• Link de acesso: http://urlib.net/sid.inpe.br/mtc-m16c/2019/10.21.13.12

Collective behaviors emerge from interactions among living beings. Similarly, engineering applications with mobile robots might employ bio-inspired strategies to reach desired group configurations. This thesis proposes strategies that exploit communication among close agents to obtain the desired behavior. Firstly, a reactive model of agents that group into a certain neighborhood and follow a mobile reference is presented. The adopted parameters are achieved via heuristics that target specific formation topologies. Each agent has a sensory region around itself that allows perceiving neighbors that come close. Furthermore, the group follows the socalled virtual agent (VA), a non-real agent that dictates the trajectory to be chased. The second part of the thesis is devoted to a first-order model of particles with coupled oscillator dynamics, focusing on the specific case of particles moving in circular trajectories and reaching clusters of the same size that are symmetrically distributed along the trajectory. The study starts with an investigation of the control parameters space in order to find regions where particles reach desired formations regardless of initial conditions. After that, the effects of adding and removing particles from already stable ensembles are explored. Simulations show that some particles never join clusters when the number of new elements is smaller than a certain critical value. In a similar fashion, clusters start to break only after a critical number of removed elements. Additionally, a strategy is proposed to change from one cluster configuration to another. Finally, the so-called Switching System is proposed, allowing for symmetric circular formations with time-periodic and chaotic dynamics. The model for obtaining cluster configurations is implemented in a simulator of mobile robots as a proof of concept.