High noise and complicated errors caused by low-cost MIMU (micro-electro-mechanical system-based inertial measurement unit, MEMS-based IMU) have caused its stochastic modeling challenge, which may undermine the performance of inertial-based integrated navigation. In order to achieve accurate MEMS-based navigation, a stochastic modeling method called auto-regressive moving-average model for low-cost MEMS-based inertial sensors was proposed on the basis of time series analysis theory. This model was then expanded into the state variables of the conventional Kalman filter to establish the system dynamic equation and observation equation and to estimate the zero-bias online. Field test results indicate that the proposed algorithm can not only realize a highly accurate autonomous navigation for lowcost MIMU, but also provide reliability to the MIMU/GNSS integrated system. 