Skip to Main Content
Proper performance of biped robots in various indoor environment conditions depends on reliable landmark detection and self-localization. Probabilistic approaches are among the most capable methods to present a real-time and inclusive solution for biped robot's localization. This paper focuses on the newly proposed odometry system implementing the kinematic model, predictive estimator of the camera position, particle filter methods such as Monte Carlo Localization (MCL) and Augmented MCL with utilization of landmarks, lines, points, and optimized filtering parameters for robots' state estimation. Moreover, kidnap scenarios which could not be considered and handled with the uni-modal Kalman filter-based techniques, are studied here for the Nao biped robots on the RoboCup standard platform league's soccer filed. Experimental data are employed to evaluate effectiveness and performance of the proposed vision module and augmented MCL scheme.