Enhancement of Heading Accuracy for GPS/INS by Employing Average Velocity in Low Dynamic Situations

Abstract

In this paper, we proposed an algorithm that improves the heading accuracy of the global positioning system (GPS)/inertial navigation system (INS) integrated system used in automobiles by manipulating INS velocity and GPS velocity measurements. Two velocities are provided by the GPS receiver: the velocity calculated using the position difference and the velocity calculated using the Doppler shift. The velocity obtained using the position difference is an average velocity for a certain time period, which is inaccurate under dynamic conditions because of its time delay. In contrast, the velocity from the Doppler shift is an instantaneous velocity and has no time delay. However, it also relies on pseudo-range error and noise, which degrades the heading estimation accuracy in low dynamic situations. Thus, although the velocity measurements can improve heading accuracy, the navigation performance is often degraded when the velocity measurements are used for GPS/INS integration. To improve the heading accuracy by solving the aforementioned problems, we proposed a heading accuracy improvement algorithm that employs the average velocity measurements obtained using the averaged GPS velocity and the average velocity of the INS. Since the proposed average velocity measurements are calculated using long baseline, the proposed algorithm can improve the heading accuracy without using other sensors, especially in the case of low dynamic situations. It can be easily applied to the existing GPS/INS integrated system, making it suitable for use in automotive navigation systems. In this research, it is verified that the average velocity measurements can be substituted for Kalman filter measurements, and the performance improvements are confirmed through simulations and experiments. INDEX TERMS Automobile vehicles navigation, GPS integrated navigation, velocity, heading accuracy, average velocity measurements.

Vehicle heading angle and IMU heading mounting angle improvement leveraging GNSS course angle

“…The course angle of vehicle and course measure flag is shown in Figure 11. It can be found that the course measure flag varies frequently from 0 to 1 according to the judgment criteria set in (12), where 0 means GNSS course angle is not used as measurement and 1 means GNSS course angle is used as measurement. From the system dynamic (6) and measurement equation (10), we can find that if no measurement is used in the Kalman filter the IMU mounting angles and vehicle scale factor will remain unchanged since they only appear in the measurement equation (10).…”

“…A simple criteria to judge whether the vehicle is traveling in straight line is shown in (12). The angular rate of IMU z axis is used to determine whether the vehicle travels in straight line, where VSFL represents the vehicle straight line flag.…”

“…Moreover, since the quality of GNSS course angle is related to the velocity of GNSS, the measurement of GNSS course angle is only used when the velocity of GNSS is greater than the threshold. For the sake of simplicity, the course measurement flag is set as the same as VSFL in (12), where v z is the angular rate of IMU z axis, v VSLF and vel th are the threshold of v z and vel th , respectively.…”

“…However, in the GNSS/INS/onboard sensors fusion framework, the attitude, specifically the heading angle, may suffer from large errors because of the weak observability. [11][12][13] This necessitates the substantial research to improve the heading angle accuracy. Except for the heading angle the IMU mounting angle is another variable which is difficult to measure directly.…”

Performance Enhancement of Consumer-Grade MEMS Sensors through Geometrical Redundancy

“…The first test aimed at assessing the improvement in attitude and position estimation granted by the redundant configuration; for this set of tests, all the available position and speed values of the GNSS have been considered. Heading, pitch and roll provided by both prototype and single SensorTile TM have been compared with the attitude estimates granted by the STIM300; in particular, results shown in Figure 12 are associated with the heading angle, i.e., the one most difficult to be estimated due to observability issues [ 38 , 39 ]. Even though the prototype outputs appeared to be slower to converge with respect to the reference IMU, it is worth noting the remarkable concurrence experienced for time intervals greater than 400 s. In particular, in the route section referred to as straight in Figure 11 , average differences between prototype and STIM300 angle estimates results as low as 0.025rad have been encountered.…”