GPS/INS integrated navigation system has been a candidate of integrity monitoring system since an inertial sensor could improve performance of the fault detection and exclusion (FDE) functions. Japan Aerospace Exploration Agency (JAXA) has developed several GPS/ INS systems called GAIA (GPS Aided Inertial navigation Avionics) for over ten years and succeeded in automatic landing of unmanned experimental vehicle in differential mode. Although high accuracy at the level of Category III approach and landing was achieved, GAIA could not be used for civil aviation since its integrity was not ensured. Therefore, JAXA has commenced research on FDE algorithms for GPS/INS navigation system, and a prototype software based on a filter bank method was developed.

On the other hand, a Japanese satellite based augmentation system, MSAS (MTSAT Satellite-based Augmentation System), which is compatible with the United States WAAS and the European EGNOS systems, has been developed by the Japan Civil Aviation Bureau (Kondo, et. al, 2001), and has been operational since September 2007. The MSAS ground infrastructure, consisting of two Master Control Stations (MCS) and four Ground Monitor Stations (GMS) located in Japan and two Monitor Ranging Stations (MRS) in Hawaii and Australia
(Figure 1). Two MTSAT satellites are in orbit as space components of MSAS.

JAXA has conducted several flight experiments since July 2007 in order
to collect the GPS/MSAS data as well as INS data for research purposes. By using these data, horizontal protection level (HPL) was computed based onthe GPS/INS filter bank algorithm, and compared to the HPL_SBAS , which was computed based on MOPS for GPS/WAAS
airborne equipment (RTCA, 2006).

The results of two flight tests are shown in this paper. One was conducted in July 2007 at Taiki in Hokkaido, northern Japan, while another test was conducted in February 2008 at Hachijo, a southern Island in Tokyo. Both locations are shown in Figure 1.

Gps aided inertial navigation avionics (gaia)

GAIA GPS/INS integrated navigation system was originally developed by the National Aerospace Laboratory (NAL) and the National Space Development Agency of Japan (NASDA) for the High Speed Flight Demonstration (HSFD) project, a test program for the planned HOPE-X
space plane (Harigae, et. al, 2001). (NAL) and NASDA, and another organization, Institute for Space and Astronautical Science (ISAS), were merged into JAXA in October 2003.) A key objective of the HSFD Phase I experiment was to examine automatic takeoff and landing technology using carrier-phase DGPS/INS (CDGPS/INS) integrated navigation, and high accuracy at the level of Category III approach and landing was demonstrated in differential mode.

Figure 1. MSAS Overview and Flight Test Area

INS (CDGPS/INS) integrated navigation, and high accuracy at the level of Category III approach and landing was demonstrated in differential mode.

The onboard avionics consists of a Kearfott T-24 Inertial Measurement Unit (IMU) with ring laser gyro and servo accelerometer, an Ashtech G12 singlefrequency GPS receiver, and a DX4 (66MHz) CPU for navigation processing. Figure 2 shows a photograph of the GAIA and Table 1 gives its specifications.

GAIA is currently installed in JAXA’s experimental aircraft Beechcraft Model 65 QueenAir and provides navigation data for research purposes. Since GAIA was not capable to decode MSAS message, an Ashtech DG16 GPS/WAAS receiver was installed for these flight tests. Onboard equipment system is depicted in Figure 3.

The data recorded onboard were used for offline analyses. The processing software was originally developed for MSASGAIA (Tomita el al., 2003) and modified to add integrity monitoring function. MSAS-GAIA is a further development of GAIA which utilizes SBAS capability. The navigation algorithm of GPS/MSAS/ INS is outlined in Figure 4. MSASGAIA adopts a tightly-coupled GPS/ INS integrated navigation algorithm that corrects IMU errors (acceleration and angular rate) as well as the INS navigation results (position, velocity and attitude) using GPS data, and avoids the divergence of inertial navigation.

Several integrity monitoring algorithms have been proposed for GPS/INS navigation system (Brenner 1995, Diesel and Dunn 1996, Young and McGraw 2003). JAXA adopted the normalized solution separation method using filter bank (Young and McGraw 2003) for a prototype software. In the analyses later, HPL are computed for en route through LNAV approach. Therefore, probability of missed alert, which is assumed to be equivalent with PMD, false alert rate, and fault-free integrity probability are 0.001, 10^-5/hour, and 10^-5/ hour, respectively. The HPL computed as above is denoted as HPL_FD hereafter.

Flight experiment and results

Two flight tests were conducted in order to evaluate the developed algorithms. The first flight experiment (CASE 1) was carried out on July 23, 2007 at Taiki (see Figure 1) in Hokkaido, northern Japan (Tsujii, et. al, 2008). The Beech 65 aircraft took off from Obihiro Airport, about 40 km north west of Taiki, flew to Taiki airfield and carried out lots of circling patterns with up to 30 degree bank above the airfield and Pacific Ocean coastline, then went back to Obihiro airport.

Figure 2. GAIA (Right, Left is an uplink receiver for DGPS)

Figure 3. Onboard Equipment System

Figure 4. Outline of GPS/MSAS/INS Navigation Algorithm

Figure 5. HPL without MSAS Correction

Figure 6. HPL with MSAS Correction

Figure 7. Magnitude of σUIVE

Figure 8. Trajectory of Aircraft

Figure 9. Roll and Pitch Angle

Figure 10. GPS/INS position error (without MSAS correction) and its estimate (95%; dotted line)

The horizontal protection level computed by the GPS/INS filter bank method (HPL_FD) and that computed by the SBAS method (HPL_SBAS)
without MSAS correction were shown in upper part of Figure 5, while the number of observed satellites and , which is a test static for the fault detection, were shown in lower part.
It is clear that integrating GPS with INS drastically reduces the value of HPL. Also, HPL_FD seemed very stable while HPLSBAS was affected by the number of satellites and resulting satellite geometry. This superior performance is basically attributed to accurate position estimate of navigation filter. HPL_SBAS was directly affected by the range error variance which was very conservative for safety reason.
On the other hand, HPL_FD was calculated based on the filter covariance, which was smooth and small due to the effects of hybridization with INS. Therefore, HPLFD was reduced even though the
same value of range error variance was used. The HPL_FD and HPL_SBAS when pseudoranges were corrected by using MSAS message were shown in Figure 6. Compared to Figure 5, HPLSBAS was improved since GPS range errors were reduced. On the other hand, improvement of HPLFD by MSAS correction was not significant. Note that there were sufficient satellites during all the flight. An example where fewer satellites were observed is shown in the next section.

The second experiment (CASE 2) was conducted 7, 2008 at Hachijo-Island, Tokyo, Japan. Since Hachijo-Island is located southern, the ionospheric effect is severe than in CASE 1. An example of vertical ionospheric delay error (σUIVE) is depicted in Figure 7. σUIVE is computed by interpolation using the grid ionospheric vertical error (σUIVE) which is given at each grid point with five degrees separation in longitude and latitude.

Trajectory of aircraft at Hachijo-Island on February 7, 2008 is shown in Figure 8. The Beech 65 took off Hachijo- Island airport and flew south east, then conducted counter clockwise circling twice and clockwise circling twice. Next, it ascended to height 1200m, and descended to 600m, then landed at the Hachijo-Island airport. The height of the airport above the ellipsoid (WGS84) is about 130m. The origin of time is at the completion of INS alignment. The attitude of aircraft is shown in Figure 9. Since roll angles at four circling were over 25 degrees, the satellite at lower elevation might be blocked by aircraft itself. The error of position estimated by the full filter and estimated error range (95%) from the filter covariance are shown in Figure 10 (without MSAS correction), and in Figure 11 (with MSAS correction). The efficacy of MSAS corrections was seen since the same satellites were used in both cases. The error range was properly estimated in horizontal direction, therefore the horizontal position estimates and covariance were able to use for HPL calculations. In this test, a ground GPS receiver was installed nearby the airport, and kinematic GPS solutions obtained from onboard/ground data were used to compute position error.
The horizontal protection level computed by GPS/INS filter bank method (HPLFD) and that computed by SBAS method (HPLSBAS) without MSAS correction were shown in upper part of Figure12, while the number of observed satellites and were shown in lower part. When six or more satellites were observed, HPL was reduced by integrating INS with GPS. At the time 2083s, number of observed satellites dropped to five and gradually increasing HPLFD temporally exceeded HPL_SBAS. During this period, HPL_SBAS did not change drastically compared to HPL_FD, which became large due to the decreased number of satellites. This is because an undetected satellite failure is assumed in computation of HPL_FD and therefore HPL_FD may become large when sufficient satellites are not observed. On the other hand, all satellites are assumed healthy in computation of HPL_SBAS if no fault is broadcasted. It was seen in lower part of Figure 12 that became very large at the time 1420s, where the number of observed satellites dropped to five temporally due to the circling with large bank angle (see Figure 9.). When number of observed satellites is five and six, the thresholds of for fault detection are 6.5286 and6.5564, respectively. Therefore, a satellite fault was falsely detected in this case.

Next, HPL_FD and HPL_SBAS with MSAS corrections are shown in Figure 13. Compared to Figure 12., it is clear that HPLSBAS was reduced by using
MSAS corrections. The HPL_FD with MSAS corrections was smaller than
the HPL_SBAS except the case where five satellites were observed. Also, the value of at about t=1420s did not exceed the threshold. MSAS corrections was effective to make the GPS/INS based FD function work properly when the number of satellites was insufficient.

Summary

An integrity function based on a filter bank method was implemented into the navigation software of JAXA’s GPS/INS avionics, GAIA. Flight
experiments were conducted and offline analyses using collected data of GPS, MSAS, and INS were carried out.

As results, when six or more satellites were observed, the HPLFD based on GPS/ INS was better than HPLSBAS, in both with and without MSAS corrections. However, if there were only five satellites observed, the HPLSBAS was sometimes smaller than HPLFD. This is because an undetected satellite failure is assumed in computation of HPLFD and therefore HPLFD may become large when sufficient satellites are not observed, while all satellites are assumed healthy in computation of HPLSBAS if no fault is broadcasted. Also, a satellite fault might be detected falsely when only five satellites were observed and MSAS corrections were not used. Therefore, MSAS corrections would be necessary in order to use the GPS/ INS based integrity monitoring when the number of satellites was insufficient.

In this paper, the GPS/INS based HPL for only en route through LNAV approach was investigated. In order to apply this method for precision approach, the vertical protection level has to be computed. However, as seen in Figure 10 and 11, the vertical positioning error could not be properly monitored by the used GPS/INS navigation filter. An improvement of vertical positioning performance by an ionospheric delay estimation method such as precise point positioning (PPP), and/
or by an integration of altimeter would be required. Future GNSS such as GPSIII and GALILEO, which will implement dual frequency operation, would make precision approach possible. JAXA plans to conduct simulation analyses to investigate the potential of future GNSS/INS for precision approach.

Figure 11. GPS/INS position error (with MSAS correction) and its estimate (95%; dotted line)

Figure 12. HPL, number of satellites, and √⎯λmax (without MSAS corrections)

Figure 13. HPL, number of satellites, and √⎯λmax (with MSAS corrections)

References

Brenner, M. (1995). Integrated GPS/Inertial Fault Detection Availability, Proceedings of ION GPS-95, Palm Springs, CA.
Diesel, J. W., and Dunn, G. (1996). GPS/IRS AIME: certification for Sole Means and Solution to RF Interference, Proceedings of ION GPS-96, Kansas City, MO.
Harigae, M., Nishizawa, T. and Tomita, H. (2001). Development of GPS Aided Inertial Navigation Avionics for High Speed Flight Demonstrator. Proceedings of the 14th International Technical Meeting of the Satellite Division of the Institute of Navigation, pp. 2665–2675.
Kondo, T., Kubo, N., Ishita, T., Hoshinoo, K. and Kawai, Y. (2001).
MSAS Status and Preliminary Performance Evaluation, Proceedings of
GNSS2001, Seville, Spain, 8-11 May.
RTCA Working Group SC-159. (2006). Minimum Operational Performance Standards for Global Positioning System / Wide Area Augmentation System Airborne Equipment, Document No. RTCA/DO-229D, December 2006.
Tomita, H., Harigae, M. and Hoshinoo, K. (2003). Flight Evaluation of GPS Aided Inertial Navigation Avionics with MSAS Augmentation (MSAS-GAIA). Proceedings of the 16th International Technical Meeting of the Satellite Division of the Institute of Navigation, ION GPS-2003, pp. 2819–2827.
Tsujii, T., Tomita, H., Fujiwara, T., and Harigae, M. (2008). Preliminary
Experiments of GPS/INS Based Integrity Monitoring Using MSAS Differential Correction Data, Proceedings of the 17th World Congress The International Federation of Automatic Control, Seoul, Korea, July 6-11, 2008, pp.4713-4718.
Young, R. S. and McGraw, G. A. (2003). Fault Detection and Exclusion Using Normalized Solution Separation and Residual Monitoring Methods, Navigation, Vol. 50, No.3, Fall 2003, pp. 151-169.