Fully integrated vision based localization in low cost robot using kinect

The goal of present work is to demonstrate the possibility to build a fully integrated low cost robot using an artificial vision based localization system known to be utilizable in many environments due to its extreme versatility. To avoid complex algorithms and to bound high hardware resources, simple landmarks are exploited to reduce considerably cost and complexity of recognition system. On this base a low cost project using on the shelf components has been developed in academic environment. Global architecture of the project is shown including a Graphical User Interface (GUI), a network system, and the robot itself. The algorithm allowing 3D vision based localization on a low performance embedded computer is also presented and results are discussed.