Localization and Mapping based on Range Only Sensors

Herranz, Fernando

Year: 2013

Type of Publication: Phd Thesis

Keywords: localization; mapping; slam; range only sensors; wifi

University: University of Alcalá

Abstract:

Mobile robotics has become a popular topic in robotics due to the large number of service problems that can solve. Due to this fact, the demand for autonomous navigation and guidance applications has been increased in the last decades. These applications are based on the localization and mapping. Localization problem is solved using maps and sensor information of the environment. On the contrary, mapping problem is solved using location and sensor information. By means of the information provided by range only sensors and probabilistic techniques, it is possible to obtain beacon maps of environment and estimate the user location. Nowadays, range only sensors are low-cost, non-intrusive sensors and provide an unique identifier per beacon but they do not provide any angle information which makes it difficult localization and mapping. In outdoor environments, navigation is a well managed problem by global positioning system. However, it is generally not suitable to establish indoor localization and mapping systems, since microwaves will be attenuated and scattered by roofs, walls and other objects. In indoor environments an accurate observation model is needed to provide a good relationship between signal level and distance. This thesis aims at obtaining a generic observation model of WiFi signal propagation. In this thesis several algorithms are proposed in order to obtain a range only-based localization from a prior beacon map. For that purpose, it is necessary to compute a beacon map of the environment previously. Lateration techniques are presented as simple solutions to the localization problem. However, their performances are affected by geometry problems and the noise that affects the range only measurements. On the contrary, probabilistic methods such as particle filter are able to manage the characteristic of range only sensors and provide an accurate localization. The map can be built using mapping techniques which are tested in this thesis. Mapping must deal with the lack of angle information of range only sensors which makes it impossible to estimate the beacon location with only one sample. Particle filter has been presented as a good choice to map the environment when having an accurate location of the robot. For computing the map when the location of the robot is inaccurate, this thesis employs Simultaneous Localization and Mapping techniques with a robot that obtains range and odometry information. Several techniques such as filtering and smoothing approaches are tested and validated. Finally, the best combination of mapping and localization techniques (smoothing and particle filter) are used in a real application to provide an effective collaboration of human- robot teams within ABSYNTHE project.