As for the special application environment of underground tunnels, an auto\|driving technique of unmanned underground scraper was proposed based on optimal trajectory tracking. The core of this method is acquiring horizontal position deviation, course angle deviation through vehicle\|sensor while the scraper is under auto\|driving in the tunnels, with deviation information being mixed to realize the tendency to zero of deviation by adjusting fore and post splice angle of scraper through real\|time control. A navigation coordinate has been set for controlling while the scraper is ready for auto\|driving firstly. Then the principle of the controlling algorithm was introduced. How to calculate the horizontal position deviation and course angle deviation was explored in detail with the laser measurement system mounted on the top of the scraper’s cab. The deviation was fused integration for the controlling data. The step of real\|time controlling algorithm was executed in order to realize unmanned driving of underground scraper with steering systematical movements controlled by output controlling voltage of the valve. Simulation test on the algorithm was carried out combined with practical application. To further verify the feasibility of the algorithm, the method was applied to the real auto\|driving control of two\|cubic diesel scraper. The experimental result shows that this control algorithm based on deviation fusion can realize the auto\|driving of scraper. The auto\|driving control system can be adjusted easily and quickly as the control system owns fast tracking speed and realizes balance between fast and smooth without great overshoot and oscillation.