Abstract: The ranging accuracy will evidently decrease in underground roadway because of multi-path effect and NLOS signal. In order to solve this problem,the similar statistical law of time delay between NLOS signal and LOS sig- nal at their receiving ends is discovered based on the deviation generation mechanism of spread spectrum ranging of mine roadway. Thus a positioning method based on the two-reconstructing ranging value for curved roadway is presen- ted. Taking fully advantage of location information of roadways,this method firstly conducts the first reconstruction on the NLOS ranging value of curved roadways in order to get the estimate of linear distance between the moving target and the corresponding base station so the observation equation with the reconstruction figure can be established. Sec- ondly,this method carries out the secondary reconstruction and projects the solution point of observation equation to roadway space so that location estimate point can be obtained. Finally,Kalman filter is used to further optimize the po- sitioning accuracy of position estimation point and hence make it possible for positioning under a more general curved roadway. Simulation performance shows that the method can obtain a desired positioning accuracy in curved roadway with mean curvature from 0 to 0. 30. The experiment,which conducted in the pedestrian curved tunnel of a subway with a mean curvature 0. 192,shows that the Root Mean Square (RMS) of positioning error is 0. 891 m and the cumu- lative distribution of the error is ideal so that it is adoptable for accurate positioning of underground personnel.