QUOTE (jmknapp @ Nov 18 2014, 04:11 PM)
![*](http://www.unmannedspaceflight.com/style_images/ip.boardpr/post_snapback.gif)
That looks promising.
Re the spin axis and CG, I take it you don't trust the ESA .obj file exactly? When I import it into an animation program (Poser Pro) I get this for the axes:
The initial touchdown point (site J) is approximately marked. This seems about right on first glance.
Is it possible the 3:1 scale discrepancy is a meters/feet thing?
I'd prefer to trust the .obj file, but not sure if it is a valid assumption. It seems the .obj file has been released to allow the public to print the comet in 3D, and I kind of suspect that the coordinate system and units may be assigned to fit the object within the working volume of most printers.
If you compute the CG based on the cloud of points making up the mesh, then you get something that is close but slightly offset from the origin. This type of analysis will definitely be erroneous anyway since the point cloud varies in density (or so it appears). I also attempted to find the principal axes by similarly computing the moments of inertia and finding the eigenvalues/eigenvectors. Attached is a set of views where the magenta circle and solid arrows show the origin of the .obj file and the principal axes computed from the point cloud, while the magenta X and dashed arrows show the computed CG and principal axes for the point cloud relative to that CG.
My location for the landing site was picked out by hand (and is simply the closest vertex in the point cloud), and has the following coordinates (w.r.t. the raw unscaled model frame):
(0.8101, -0.3011, 0.2108)
If we could come up with a more precise position in the coordinate frame of the comet model then I can improve the analysis -- JM, I'm guessing you also picked the landing site by visual inspection?
That meter-feet scaling might be the proper scale factor (about 3.28 feet per meter). My scale is determined by finding the radial location of the lander from the SPICE data, which is 2.3655 km (not to be confused with the radius of the circle that it forms while rotating with the comet). The radial location of the aforementioned landing point is 0.8896 [units of the model], so therefore the model scale should be 2.3655/0.8896 = 2.6591. Of course this is simply to make the visualization look correct, and will change if we get a more precise estimate of the landing point.
QUOTE (aholub @ Nov 18 2014, 04:53 PM)
![*](http://www.unmannedspaceflight.com/style_images/ip.boardpr/post_snapback.gif)
IMHO the axis of rotation in the Brian's video corresponds well to the ESA model.
I'm happy with the solution for now, but would really like to get the true rotation. There is a free variable in the solution that could be fixed by minimizing the offset between the estimated largest principal axis and the spin vector for Philae telemetry.