Localization of holonomous mobile robot HOLBOS using extended Kalman filter (EKF) and robotic vision

Determining the position of a mobile robot in every time instant from sensor data is the fundamental problem in mobile robotics. This paper considers a localization of holonomous mobile robot solved in this paper using two different approaches: odometry localization and landmark based localization. In both cases the robot is placed in known environment with landmarks whose coordinates were also known. Detecting the landmarks was done by using the Microsoft Kinect camera. For odometry localization four encoders were used. Data acquired from encoders and camera is fused together employing extended Kalman filter in order to get more accurate estimation of position and orientation. Obtained experimental results prove that using encoders without any additional measurements is not enough for getting reliable estimation of robots position. Odometry localization produced an error that accumulates over time, while in the case of landmark based localization, the error is kept inside acceptable limits.