[Previous] | [Session 57] | [Next]
G.P. Guizzo (Carlo Gavazzi Space), F. Angrilli (CISAS ``G.Colombo"), I. Vukman (University of Padova)
This generation of lander navigation system was born in order to cover a wide spectrum of soft landing scenarios on planets like Mars, Mercury or the Moon; it is also particularly useful when there is no a priori knowledge of the ground. The navigation system studied here is completely autonomous and able to land on various kinds of hazardous terrains with many unevennesses like mountain tops, bottoms of illuminated craters, valleys or small plateaus. In order to choose an adequate site for landing the navigation system uses stereo image pairs, obtained with a single camera at high altitudes and with two cameras at lower ones, to produce a digital elevation model (DEM) of the terrain. It uses dense disparity maps, made from the sign-of-laplacian-of-gaussian (SLOG) of the images, as input for the vertical locus method. This technique takes the advantage of a fast and easy processing of the image since it is possible to implement the required algorithms in a efficient way by the use of an ASIC. A DSP is then used for the remaining software operations, i.e. piloting and guidance of the lander. Another advantage of this method is that, unlike most methods involving active sensors, it can be used even with the camera at great distances to the target (e.g. at the beginning of the approach phase) since it is limited only by the camera field of view and resolution. It moreover doesn't rely on any special features of the terrain, like craters, rocks or other landmarks, but it is sufficient to have an illuminated and slightly textured terrain. The navigation system validation was done using a synthetic terrain generator which allowed the closed-loop simulation of the entire system.
The author(s) of this abstract have provided an email address for comments about the abstract: guizzo@dim.unipd.it