Skip to Main Content
This paper presents a mobile robot localization system for an indoor environment using an inexpensive sensor system. The extended Kalman filter (EKF) and the particle filter (PF) is used for sensor fusion in pose estimation in order to minimize uncertainty in robot localization. The robot is maneuvered in a known environment with some visual landmarks. The prediction phase of the EKF and the PF are implemented using the information from the robot odometry whose error may accumulate over time. The update phase uses the Kinect measurements of the landmarks to correct the robot's pose. Experiment results show that, despite its low cost, the accuracy of the localization is comparable with most state-of-the-art odometry based methods.