Abstract: Autonomous vehicle was considered to be one of the toughest challenges in the fields of artificial intelligent, pattern recognition, computer vision, sensor fusion and control theory. In the paper, a kind of scaled-down autonomous vehicles was demonstrated and a more human-like environment perception and control system were proposed, including obstacle detection, combined feature-based and model-based lane mark detection algorithm. To verify the efficiency of the proposed perception and control system, a test platform was built. With the platform, the interaction behaviors between autonomous vehicles were studied.