An intelligent behavior control system for an autonomous mobile robot operating in an unstructured environment with sensor uncertainties is proposed. This study focuses on implementing and improving the methodology from Motlagh et al. on a two-wheeled P3DX mobile robot. Motlagh et al. verified their design with computer simulation. When applying it on a real robot platform, we noticed some problems and improved the design using sensor selection strategy, safe rule and target switching strategy. The proposed sensor fusion architecture introduces two additional sensors, a laser range finder on the robot and an omnidirectional CCD camera on the ceiling, to improve the reliability of the sensing capability of P3DX mobile robot. The target switching strategy is used to guide the robot out of a dead zone and reach the target by creating a virtual target. Results of the experiments with the U-shaped and subspace dead zones are presented. These results proved that the target switching strategy successfully dealt with the loop path problem in the dead zone.