Integrated GPS/INS can be implemented using a Kalman filter in different modes, such as loosely, tightly and ultra-tightly coupled.

In these integration modes, the INS sensor error states, together with all navigation error states and other unknown parameters of interest, are estimated using a dynamic model and GPS measurements such as Doppler, pseudo-range, and/or carrier phase. In order to obtain high accuracy of positioning results using such systems, the carrier phase measurements have to be used in the integration filter update. Although integrated GPS/INS systems using carrier phase observations have been developed for high precision surveying applications, most of the systems have been implemented using expensive dual-frequency GPS receivers and/or a navigation-grade INS. SNAP has been investigating the development of a cost effective GPS/INS integration using a pair of single-frequency GPS receivers and a tactical-grade Strapdown INS (SDINS), which delivers centimetre-level positioning accuracy even during a few seconds of GPS signal blockage. Figure 1 depicts the GPS/INS integration scheme used.

Figure 1: Tightly coupled GPS/INS integration scheme using carrier phase measurements

On 24th and 25th March 2003, kinematic experiments were carried out in the Clovelly Bay Carpark, Sydney. Both the INS and the GPS antenna were mounted on the roof of the test vehicle (see Figure 2). For data acquisition, raw INS sensor measurements were recorded at 100Hz, while GPS data were logged at 1HZ. The objectives were: (a) to evaluate overall performance of the GPS/INS integration consisting of a tactical-grade INS and single-frequency GPS receiver, under benign and harsh (signal blockage) operational environments; and (b) to investigate the impact of vehicle dynamics on integration filter initialisation and system performance during GPS signal blockage.

Figure 2: Trial vehicle

The graphs in the first column of Figure 3 show navigation solution obtained by the integrated GPS/INS system, whereas the plots in the second column depict RMS errors of the estimated navigation parameters. On the other hand, Figure 4 presents the system performance during fifty second signal blockage. Graphs in the first column illustrate RMS errors of the INS-predicted antenna position, whereas those in the second column depicts the INS-predicted antenna position accuracy obtained from comparison with positioning results without the signal outages.

Figure 3: Navigational parameters and their RMS errors

Figure 4: INS-predicted antenna position errors during 50 seconds of the signal blockages

In order to investigate the influence of vehicle dynamics on the estimation of the error states, four experiments were carried out with controlled-trajectories. Figure 5 shows the RMS errors of the horizontal accelerometer biases and heading error estimation, indicating the different vehicle dynamic contribution to the Kalman filter estimation procedure. These test results suggest: (a) vehicle dynamics affect the Kalman filter initialisation time and estimation performance, especially the heading component; (b) the higher the dynamic changes in the lateral direction, the shorter the initialisation time and the more precise the filter estimation; and (c) the S-turn shaped trajectory provided the best system performance among the four trajectories considered in these tests.

Figure 5: RMS errors for horizontal accelerometer bias and heading error estimation