2-4-4 
Figure 4. Longitude difference between solutions with 
correct and incorrect lever arm components 
and no maneuvering at the beginning of the 
trajectory (10~ 6 deg corresponds to ~ 10 cm). 
GPS time of week [sec] 
Figure 5. Height difference between solutions with 
correct and incorrect lever arm components 
and no maneuvering at the beginning of the 
trajectory. 
2.2. Airborne tests in the presence of strong GPS 
interference 
Most of our test flights up to date were conducted over 
rural areas, where the radio interference generally does 
not pose any problems for the GPS signal acquisition. 
The recent flight over MIT, however, provided entirely 
different experiences, as GPS losses of lock were very 
frequent and excessively long (up to 20 minutes), which 
did not allow for a reliable INS bridging. The free 
navigation solution error grew up to hundreds of meters 
during the longest GPS gaps, preventing us from using 
the direct orientation for these parts of the trajectory. 
Therefore, the ground control points had to be 
established and surveyed in order to provide an 
aerotriangulation solution for those images that did not 
have accurate direct orientation available. 
Several flights were conducted in December 1998 over 
the MIT campus with the objective of collecting 
directly oriented imagery (for an example image see 
Figure 6) in order to build a 3D model of the campus 
area. Two base stations, at Norwood Airport and at 
MIT campus, collected GPS data during the flights. The 
flight conditions were very different from those 
experienced during the flights over primarily rural 
areas. The level of signal interference over the campus 
area was so strong that lock to GPS was lost several 
times during every flight. Only early morning hours 
allowed for receiving a reasonably good and strong 
GPS signal by the rover receiver. Still, the MIT base 
station was able to collect only LI data for most of the 
mission duration. As an effect of this scenario, several 
portions of the trajectories were obtained from the free 
navigation solution, with quality depending on the total 
duration of rover GPS loss of lock, and the time of INS 
calibration before the loss of GPS signal. Table 1 
presents the error growth in latitude, longitude and 
height (only forward direction) during the GPS losses of 
lock, with the total length of the INS calibration period 
preceding the GPS gap. The difference is shown at the 
end of the gap, where the reference GPS trajectory was 
available. 
Figure 6. Aerial image of the MIT campus. 
For the quality test, the trajectories were estimated 
independently from two base stations separated by 
about 21 km. The trajectory for the test flight on 
December 12 is presented in Figure 7, and the 
differences in latitude and longitude are presented in 
Figure 8, as an example of the typical results. Assuming