Simultaneous Localization And Mapping for robots: an experimental analysis using ROS

The field of robotics has seen major improvements in the past few decades. One of the most important problem researchers all around the world tried to solve is how to make a mobile robot completely autonomous. One important step to achieve this goal is to create robots that can navigate an unknown environment and, using several sensors, build a map of it, locating themselves on the map. This particular problem takes the name of Simultaneous Localization And Mapping (SLAM) and it is very important for different scenarios, such as a mobile robot that navigates an indoor environment, where GPS location cannot perform well. Theoretically, this problem can be considered solved, since several solutions have been proposed in the literature, but in practice these solutions perform sufficiently well only under particular condition, such as when the environment is static and its dimension is limited. In real world instead, the environment can change, objects can be moved, and external factors can modify the appearance of a place, making the localization of the robot very uncertain. Therefore, in practice, a long-term SLAM is an unsolved problem and it is an open field of research. A practical problem for which a definitive solution hasn’t been proposed yet is the Loop Closure Detection (LCD) issue, that is necessary to achieve a real long-term SLAM, and it is the ability of the robot to recognize places previously visited. There are many solutions proposed in the literature, but it is very challenging for a robot to recognize the same place at different time in the day, or in different seasons, or again when the particular location is not visited for long time. During the years, several practice SLAM solutions have been implemented, but a really long-term SLAM hasn’t been reached yet. In this thesis a comparison is made between two mature SLAM approaches, highlighting their criticalities and possible improvements in view of a long-term SLAM.