Optimização Topológica

This dissertation approaches the localization and navigation problems for a differential mobile robot. Thus, the robot's architecture was updated, so that it could use the local localization system based on odometry. It was also developed a global localization system based on vision in order to correct the local localization system, iGPS (indoor Global Positioning System [Christo et al. 2009]). To identify the robot in the YCbCr color space it was used two targets and the statistical Mahalanobis distance. A Kalman filter (KF) did robot's position tracking. Intrinsic camera parameters were obtained [Bouguet 2008]. Robot's localization was corrected by a Kalman filter (KF), which combines the information from both localization systems. Robot and iGPS were integrated in a service oriented architecture through the development of an interface between Matlab® and Web Services. A proportional integral (PI) and a time-varying non linear controllers were presented as solutions to the navigation problem. Three trajectories were defined for acquiring simulation and experimental results, in order to validate the proposed methodologies. Experimental results shows that the relocaliztion system can indeed correct the odometry error, while not destabilizing the robot's navigation.