| dc.description.abstract | Humanoid robots, such as Softbank’s Pepper, are mainly used as eye-catcher during events. Businesses would like to see such robots being used more functionally. They could be used as receptionist or guide. To accomplish these tasks, the robots need to navigate within an environment robustly. This thesis aims to determine whether Pepper, in combination of existing algorithms, is suitable to create maps, localize and navigate successfully. The research question is as follows: Is Pepper suitable for autonomous exploration, Simultatenous Localization And Mapping (SLAM) and navigation using Robot Operating System (ROS) without the usage of external sensors? This research question is answered by three experiments. The SLAM experiment has shown that Gmapping can use the 3D sensor to create accurate maps. The second experiment has shown that Pepper can have difficulty localizing itself, causing undesired navigational behaviour. This is caused by the narrow field of view and the relative low quality of the 3D sensor of Pepper. The last experiment has shown that, mainly due to the failure of localization, Pepper cannot autonomously and completely explore an unknown environment. The results of the experiments indicate that, using the algorithms as defined in this thesis, Pepper is not suitable for autonomous exploration, SLAM and navigation. Further steps could be undertaken to improve these results. Better sensor hardware or adaptations in the algorithms could make the difference in localization and make Pepper suitable for these tasks. Improving the maps manually or automatically, or using landmark detection are also interesting methods that could improve the behaviour of Pepper. | |
