Autonomous Navigation in Dynamic Environments

Abstract

This thesisstudies in details the modular architecture to designing autonomous mobile robotswhich has four different layers: perception, localization, planning and control. This architecture aims to provide a test bed for researchers wherethey can rapidly prototype the application of new methods without the effort of redesignthe entire infrastructure.The document also contains a background about the relevant topics and related works reviews, adetailed description of the approach taken, tests and the consecutive results. It discusses an implementation of the modular architecture in a quadcopter platform to convert it from an ordinary human-controlled quadcopter to an autonomous quadcopter that can navigate the environment without any human aid. Different algorithms are used on the different layers such as attitude planner and attractive and repulsive algorithm in the planning layer in addition to the cascaded PID regulator in the control layer. A ground station is used to provide the interface between the user and the platform for monitoringpurposes. Virtual Robot Experimentation Platform(V-REP) is used to simulate the behavior of the autonomous quadcopter in an indoor environment that contains some obstacles