Abstract: Based on Dempster Shafer theory of evidence inference, this paper presents some algorithms of multisensor data fusion for mobile robots. By integrating the 2D data and the 3D data that are both the pre processed information obtained by a color CCD camera and a laser imaging tange sensor respectively, the accuracy and the reliability of environment understanding and obstacle detection for mobile robot are improved than those based only on any single sensory information. Some difficult problems of vision information fusion for mobile robot are discussed, and meaningful expermental results of fusion on unstructured road network are obtained.