Autonomous exploration and mapping with mobile robots

Abstract

The discovery of uncharted territories has been one of the greatest and earliest fascinations of mankind. Exploration of unknown environments mostly leads to the creation of a map of the terrain being explored. Robotics finds itself in the midst of the necessity for exploration and mapping. Until recent years, robotic mapping of environments had been performed by manually steering the robot around the environment while it uses its onboard sensors and algorithms to construct a world model. Novel research has revealed techniques through which both the robot navigation during exploration, and the mapping process are performed autonomously. In general, an autonomous exploration and mapping robotic system consists of three main components: Simultaneous Localization and Mapping, SLAM, explo ration and a path planning (motion control) component. SLAM is a process through which the robot creates a model of the physical environment (mapping) while at the same time estimates its location within that map (localization). Moreover, given a current robot location in the map, an exploration strategy decides upon the next best location in the environment that the robot should visit in order to expand its charted territory and improve localization. Finally, the path planning and motion control component is responsible for the safe planning and execution of a path for the robot from its current location to a goal location. This dissertation aims to investigate the state-of-the-art techniques that can be used to design and implement such a robotic system. For this purpose, the research robot, Powerbot™, equipped with Robot Operating System (ROS) - a software development framework for robots - was used. The modularity of the implemented scheme enabled three different exploration strategies to be experimentally validated on Powerbot™, running the same SLAM, path planning and motion control components, in a real-life environment. This work contributes to the robotics community by providing ROS implementations of three exploration strategics. In addition these strategies are validated and compared on a real robot running ROS. The experimental results show that there is a statistically significant difference between the performance of the three strategies. These differences are analysed and discussed in detail in this dissertation.