An Estimation Method of Vehicle Position for Automated Driving with GNSS

Accuracy of positioning with GNSS (Global Navigation Satellite System) has been improved in recent years. Especially in Japan, high accuracy GNSS service, QZSS (Quasi Zenith Satellite System), will start in 2018 and the first QZS, “MICHIBIKI” has been already launched. They will broadcast correction data which enhances GNSS performance and realize cm-order positioning. In this paper, we, Mitsubishi Electric develop the estimation algorithm of vehicle position and attitude and also adapt the algorithm to a test vehicle which can trace automatically the calculated path with EPS (Electric Power Steering) and high accuracy GNSS. Although the GNSS receiver calculates the longitude and latitude of the vehicle every second, it is not enough to control vehicle dynamics smoothly. So we estimate vehicle position and attitude of the vehicle with GNSS and vehicle sensors in high frequency. Using them, we calculate the target path as cubic function from road dot sequence in a body fixed coordinate. Target steering angle is decided with a look-ahead driver model and we control the EPS to turn the tires to trace the path. We test the automated driving vehicle which uses EPS, road map and GNSS with correction data from internet instead of QZSS. It can trace automatically the target course in our test course and achieve sub-meter accuracy.