Localization and Mapping for Service Robotics Applications

In this work we address different techniques for autonomous navigation of single and multi-robot systems in unstructured indoor environments. The first problem that we address is the problem of robot localization. We first show how a multi-robot localization approach can speed up the process as well as introducing robustness to failures. At the same time we show the possibility to introduce some sort of insight into the algorithm. We then focus on increasing the robustness of single robot localization, by taking advantage of sensor fusion and by exploiting visual markers embedded in the environment. We then show how the approach can be extended from ad-hoc markers to an arbitrary set of pre-existing objects already present in the environment. Then we address the more complicated problem of Simultaneous Localization and Mapping. In particular we focus on a promising technique called Graph-SLAM and we show the implementation of a front-end for this optimization technique in order to create a real online SLAM approach. The last part of the thesis is devoted to another part of service robotics which is not directly related to navigation: Human Robot Interaction (HRI). While this field of robotics is not strictly related to autonomous navigation, some part of it can be, such as the case of user following and object tracking. In Chapter 4 we discuss the implementation of an adaptive people tracking algorithm tailored for user tracking and following. We also show how this algorithm can easily be extended to the case of generic object tracking for human-robot cooperation.