AGV is one of important conveying tools in advanced manufacturing system. In order to convey material to the destination safely, exactly and real-timely, the controlling system of AGV must have some functions such as perceiving the surrounding, planning the path, positioning and controlling. Positioning is the key function for AGV to convey the material real-timely and accurately. In this paper, the novel Distributed Kalman Filter based on multi-sensor was designed for XAUT·AGV100 investigated. AGV's attitude angle was estimated by fusing the metrical information from sensors such as encoders, gyro and ultrasonic sensors. It was indicated by the simulation that the designed filter is reliable and effective. It is tested by the experiment of AGV tracking a linear-path that the filter can estimate AGV's attitude angle accurately and timely while AGV is moving. The positioning accuracy of the filter can meet the need of positioning AGV in manufacturing system through theory analysis and experiment.