Programa de Pós-graduação
em Engenharia Elétrica


Defesa Pública da Dissertação de Mestrado do Engenheiro de Computação RODRIGO SAAR DE MORAES.

Data: 27 de Março de 2018  - terça-feira
Horário: 13h30min
Local: Sala 301-A do Instituto Eletrotécnico (Av. Osvaldo Aranha - 2º andar)

Banca Examinadora:
Prof. Dr. Luís Álvaro de Lima Silva - PPGI-UFSM
Prof. Dr.  Bruno Castro da Silva - Deptº Informática Aplicada - Inst. Informática - UFRGS
Prof. Dr. Walter Fetter Lages - PPGEE-UFRGS
Prof. Dr. Édison Pignaton de Freitas - PPGEE-UFRGS (Orientador)


Abstract: "Observing the current scenario, where terrorism and vandalism acts have become commonplace, particularly in big cities, it becomes clear the need to equip law enforcement forces with an efficient observation method, capable of identifying and observing potentially threatening individuals on crowds, to avoid or minimize damage in case of attacks. Moreover, with the popularization of small lightweight Unmanned Aerial Vehicles (UAVs), these have become an affordable and efficient tool, which can be used to track and follow targets or survey areas or buildings quietly, safely and almost undetectably. This work presents the development of a multi-UAV based crowd monitoring system, demonstrating a system that uses small Commercial Of The Shelf (COTS) UAVs to periodically monitor a group of moving walking individuals. The goal on developing this work was investigating how well agroup of UAVs can continuously monitor a large group of individuals (targets) in a crowd, alternately visiting each of them at a time while trying to not lose sight of any of these targets. A system equipped with a group of UAVs running this proposal can be used for law-enforcement applications, assisting authorities to monitor crowds in order to identify and to follow suspicious in dividuals that can have attitudes that could be classified as vandalism or linked to terrorist attack attempts. To address this problem a system composed of three parts is proposed and was developed in this thesis. First, an auction algorithm was put in place to distribute interest targets among the multiple UAVs. The UAVs, in turn, make use of a genetic algorithm to calculate the order in which they would visit each of the targets on their observation queue. Moreover, a target handover algorithm was also implemented to redistribute targets among the UAVs in case the system judged that a target was about to be lost by its current observer UAV. The proposed system was evaluated through a set of experiments set-up to verify and to demonstrate the system capabilities to perform such monitoring task, proving its efficiency. During these experiments, it is made clear that the system as a whole has a great potential to solve this kind of moving target monitoring problem that can be mapped to a Time Dependent Travel Salesman Problem (TDTSP), observing targets, and redistributing them among UAVs as necessary during the mission.

 Keywords: Multiple-UAV systems, target tracking, crowd monitoring, inteligent control system, UAV monitoring, Multi-target observation"