In order to improve the performance of a GPS/DR integration system, the error sources of DR sensors should be modeled accurately, This results in the increases in the dimension of the integration filter and, consequently, computational load becomes large. To reduce the computational load, suggested in this paper is a decoupled GPS/DR integration scheme that consists of two cascaded Kalman filters. The GPS velocity output is used in the first filter to calibrate the DR sensor and to fix the velocity as well. The velocity from the first filter is fed to the second filter where the position is corrected using the GPS position output. Experimental results show that the proposed integration scheme has positioning performance comparable to the conventional coupled one, while its computation is reduced to about 2/3.