Recursive estimation of vehicle position by using navigation sensor fusion

In this paper, a sensor fusion scheme is employed to reduce positioning error of a vehicle since the GPS signal is fail. The vehicular information, such as position, heading direction, and velocity, can be obtained through GPS signal. Generally, the positioning accuracy of commercial GPS module is within the 3 meters, however, the GPS module may disconnect the signals from satellites since the vehicle is maneuvered under shelters, e.g. parking garage, tunnel, high dense urban, etc. Therefore, our proposed methodology is able to improve the estimation accuracy of vehicle position based on dead reckoning method. The first step, the Kalman filter is utilized to reject the noise of velocity measurement which is captured from gearbox and wheel speed sensor and also predict the velocity and displacement of vehicle in next sample time. The second step is to construct the displacement model of the vehicle by adopting ARMA model, which is able to estimate the state of vehicle. Digital map information which is applied to correct the positioning result of ARMA model is addressed in the last step. A real time experiment result of GPS signal lost in a tunnel is carried out to demonstrate the performance of our proposed method.