|Full text PDF:
The idea of having robots performing the task for which they have been designed completely autonomously and interacting with the environment has been the main objective since the beginning of mobile robotics. In order to achieve such a degree of autonomy, it is indispensable for the robot to have a map of the environment and to know its location in it, in addition to being able to solve other problems such as motion control and path planning towards its goal. During the fulfillment of certain missions without a prior knowledge of its environment, the robot must use the inaccurate information provided by its on-board sensors to build a map at the same time it is located in it, arising the problem of Simultaneous Localization and Mapping (SLAM) extensively studied in mobile robotics. In recent years, there has been a growing interest in the use of robot teams due to their multiple benefits with respect to single-robot systems such as higher robustness, accuracy, efficiency and the possibility to cooperate to perform a task or to cover larger environments in less time. Robot formations also belongs to this field of cooperative robots, where they have to maintain a predefined structure while navigating in the environment. Despite their advantages, the complexity of autonomous multi-robot systems increases with the number of robots as a consequence of the larger amount of information available that must be handled, stored and transmitted through the communications network. Therefore, the development of these systems presents new difficulties when solving the aforementioned problems which, instead of being addressed individually for each robot, must be solved cooperatively to efficiently exploit all the information collected by the team. The design of algorithms in this multi-robot context should be directed to obtain greater scalability and performance to allow their online execution. This thesis is developed in the field of multi-robot systems and proposes solutions to the navigation, localization, mapping and path planning processes which form an autonomous system. The first part of contributions presented in this thesis is developed in the context of robot formations, which require greater team cooperation and synchronization, although they can be extended to systems without this navigation constraint. We propose localization, map refinement and exploration techniques under the assumption that the formation is provided with a map of the environment, possibly partial and inaccurate, wherein it has to carry out its commanded mission. In a second part, we propose a multi-robot SLAM approach without any assumption about the prior knowledge of a map nor the relationships between robots in which we make use of state of the art methodologies to efficiently manage the resources available in the system. The performance and efficiency of the proposed robot formation and multi-robot SLAM systems have been demonstrated through their implementation and testing both in simulations and with real robots.