engleski

Pose estimation of mobile robots in partially and completely unknown environments

This disertation investigates algorithms for mobile robot's pose estimation in known and unknown environments. Probabilistic approach to mobile robot's pose estimation is used because environment model, mobile robot's dynamics model and sensors measurements are stochastic in nature. New algorithm for estimation of mobile robot's absolute pose is developed using RF ultrasound positioning sensors which, besides mobile robot's position, estimates also its orientation. Algorithm for relative mobile robot's pose estimation with monocular vision in known complex indoor 3D environments is also developed with special emphasis on modelling those environments, realtime execution in image features extraction and 3D model rendering. Robust simultaneous localization and map building (SLAM) algorithm based on stereo vision is developed. Its properties from estimation theoretical and implementational aspect are considered. All the algorithms are tested by simulations and experimentally.

mobile robotics; pose estimation; RF ultrasound sensors; robot vision; SLAM

nije evidentirano