Robot visual navigation using ceiling images

The article reviews the operation support of on-board navigation system, which locates robot indoors.Navigation system operational errors caused by (accumulated) accelerometer, angular velocity and encoder errors should be compensated by updating a robot position. The navigation system update involves various methods, such as visual navigation executed by on-board synthetic vision facilities.In major methods applied, the synthetic vision-based navigation system can operate in visual line of sight. Among disadvantages of methods using special markers is them being necessarily installed in various indoor positions (floor, ceiling), that requires extra time and resources to arrange the needed data field. Besides, such methods turn out to be ineffective in dynamically varying situations like in crowded rooms, where moving people or robot vehicles may block landmarks and disrupt the navigation system.The method offered by the authors reviews the integration of navigational data received from encoder and synthetic vision system, and based on Extended Kalman Filter (EKF). The position can be found with the synthetic vision system through the current ceiling sections images obtained by a robot being referenced to the prearranged image of the whole ceiling. The referencing is executed with an image matching correlating algorithm. The offered system helps accomplish the robot navigational tasks in dynamically varying situations such as people moving around the robot. The algorithm experiments have shown the possibility of high-precise robot positioning in rooms at mean-square measurement errors less than 0,025m.