This paper builds a system to perceive the unknown environment based on multi-sensor data fusion. It enables the self-determined mobile robot to identify the type of the obstacles around it timely in the process of travel, and the mobile robot achieves more intelligent operations. First, the paper introduces the theory of the mobile robot's environmental perception and the configuration of the multi-sensor data fusion technology. Second, it combines the characteristics of neural network (NN) and brings forward that apply the MLP (multi-layer perception), which based on the improved back propagation (BP) arithmetic and four mature items of fusion rules, to perceive the real surroundings aiming at simplifying the former solving projects. Third, the paper presents a structure of multi-sensor data fusion, which two sub-networks for barrier recognition are separately built and incorporates with each other by parallel connection to form a high-powered recognition system. Finally, the results of experiment present that the method given by the paper are very pleasant