This paper presents the implementation of a perceptual system for a mobile robot. The system is designed and installed with modern sensors and multi-point communication channels. The goal is to equip the robot with a high level of perception to support a wide range of navigating applications including Internet-based telecontrol, semi-autonomy, and autonomy. Due to uncertainties of acquiring data, a sensor fusion model is developed, in which heterogeneous measured data including odometry, compass heading and laser range is combined to get an optimal estimate in a statistical sense. The combination is carried out by an extended Kalman filter. Experimental results indicate that based on the system, the robot localization is enhanced and sufficient for navigation tasks.