7.12.16 - Conclusion of the experiment: Final autonomous approach to 30m

Based on the confidence gained during the first autonomous approach, a second rendezvous has been exercised during the few remaining days of the experiment. The objective was to explore the behavior of the autonomous vision-based GNC system at mid to close range. The main difficulty here consists in tuning properly the GNC parameters to cope with the peculiarities of the close approach:

At this distance, the target is so bright that no star is visible anymore in the picture. As a result, it becomes impossible to determine precisely the orientation of the camera using the celestial objects present in the image background. The onboard navigation has thus to rely on the onboard estimate of the spacecraft attitude, which can be affected by a large error (up to several degrees) depending on the availability of the other star tracker. This additional source of error has to be reflected in the filter measurement noise: even if the target centroiding performance is still at the subpixel level (1 pixel corresponds here to 80’’), the onboard attitude error leads to a global measurement noise of about 1 degree (3600’’), which is more than one order of magnitude larger than what we were used to see at far range.
The onboard safety mechanisms have to be tuned with care to allow such a close approach. As already mentioned in a previous post, the complete AVANTI experiment relies on a minimum separation between the spacecraft which is always guaranteed. The onboard safety monitoring task declares the relative motion as too risky if this value drops below a user-defined threshold. This threshold is set considering several sources of uncertainty: onboard relative navigation errors, numerical errors of the onboard calculation, maneuver execution errors. In order to decrease the intersatellite distance to 30m, the minimum allowed distance had to be decreased accordingly after a cautious analysis of the onboard uncertainties.

Being properly tuned, the vision-based autonomous rendezvous software could perform successfully a last autonomous approach from 3km to only 30m in two days! With our formation configuration, the minimal distance was unfortunately reached during eclipse, so that we could take images up to 57m distance. Nevertheless these are pretty impressive images!

Final approach from 200m to 30m (dL stands for the relative mean longitude)

Having reached all our experimental goals, it is time for us to conclude this experiment and say goodbye to BEESAT-4!


Left: Estimated distance during the closest approach Right: Images from 200m to 57m (note that the antennas are perfectly visible at 67m distance)