Other features that differentiate our work are (i) the use of homomorphic filtering to remove lighting effects in the images, (ii) the formalization of a method to describe the shape difference between the built map and the original grid, by means of removing the scale, rotation and reflection effects and (iii) the study of the dependence of the mapping process and the http://www.selleckchem.com/products/VX-770.html resulting map against the most interesting parameters of the process.Once the map has been built, it is necessary to test if the robot is able to compute its pose (position and orientation) within the map with accuracy and robustness while knowing its location is crucial for an autonomous agent, since the pose is needed for a precise navigation. The Monte Carlo Algorithm has been extensively used in localization tasks in the field of mobile robotics, demonstrating robustness and efficiency [21].

Different approaches have been developed depending on the nature of the sensor Inhibitors,Modulators,Libraries installed on the robot. For example, Thrun et al. [22] Inhibitors,Modulators,Libraries use a laser range sensor, Dellaert et al. [23] a camera pointing to the ceiling and Gil et al. [24] a Inhibitors,Modulators,Libraries stereo camera. The information these systems provide is used to weight the particles and estimate the position of the robot. It is also possible to use external sensors to localize the robot, as Pizarro et al. [25] do with a single camera attached at a fixed place outside the robot. However, these approaches are not applicable to large configurations of the Inhibitors,Modulators,Libraries environment.In this paper we propose to solve the localization problem using omnidirectional images and global appearance-based methods, as we do in the mapping task.

Concerning the Monte Carlo localization methods using global appearance information, some similar works can Dacomitinib be found in the literature, as the one of Menegatti et al. [10], who use Monte Carlo localization with Fourier Signatures as global image descriptors and path-based maps and mainly centers in the strategy to solve the robot kidnapping problem. In our approach, we use dense maps (grid-based maps) to carry out the experimentation and we propose and compare different weighting methods to optimize the localization task. In some of these weighting methods we have included some information from the orientation extracted from the omnidirectional images, which gives robustness to these approaches.

We carry out the selleck experimentation with different grid sizes in the map and different number of particles and all the results are decomposed in global localization and tracking. The main contribution of our work in this field consists of optimizing the parameters of the particle filter, as we show in the results section.Another related work is presented by Linaker and Ishikawa [26]. They introduce PHLAC (Polar High-order Local Auto-Correlation) to describe the images in the global appearance domain, with an adaptation that makes it invariant against rotations when working with omnidirectional images.