Abstract: A new algorithm for computing traversability of search and rescue robot is presented based on terrain prediction and correction to realize autonomous navigation.The traversability in this paper includes pre-traversability and real-time traversability.The pre-traversability is obtained by calculating terrain roughness and waviness in front of the robot with image processing methods,and the robot will follow a trace with maximum pre-traversability.During the running time,the posture variations of robot and the slip between the track and ground are measured to correct the robot navigation errors caused by pre-traversability.The revolving ultrasonic-rangefinder is used to get the distance information and avoid influence of obstacles on texture information.The field experiment demonstrates the effectiveness of this navigation algorithm.