My attempt to crudely register the two OSIRIS frames and the ROLIS 40 m frame:
Click to view attachment
It's pretty clear we had 4 points of contact.

Here's a version showing the two OSIRIS frames annotated with the contacts circled and other changes marked with arrows:
Click to view attachment
The three white circles roughly match the separation of the legs in the OSIRIS approach image.

I'm proposing that the first contact was at the leftmost white circle with the leftmost (L) leg. The L leg then bounced up (but lander centre of mass (COM) continued downwards), which pivoted the lander so that next only the two right legs touched down, forming the right white circles. The right legs may have dragged across the surface a bit.

This second contact bounced the right legs up, which pivoted the lander oppositely to the previous pivot, so that even though the COM was now headed upwards, the L leg made contact one more time at the black circle.

Of course some detailed modelling would be needed to decide if this is a physically feasible explanation...

As a starter - here's a 3D Philae generated by using http://blogs.esa.int/rosetta/files/2014/11...ll_Philae_2.png and http://www.esa.int/var/esa/storage/images/...-angle_view.jpg

Click to view attachment

I missed the leftmost impact... Last input from me.


You're right, it does fit. Now to imagine how it could make the mark on the right without seeing the accelerometer data.
There is also a flywheel in the mix. Still unintuitive to me, I need to re-read how they respond to rotations.

And for fun - not supposed to represent the actual location of Philae at any time.....

They also have the navcam sighting of the orbiter a minute and a half after first bounce. Now that we have a definite OSIRIS sighting 9 minutes after first bounce, it's easy to see that the distance from the bounce site to the navcam candidate relative to the distance to the OSIRIS post-bounce location is roughly similar to the time delay ratio between them. And the directions also match well.

In other words, it now looks very likely that the navcam sighting is real.

In terms of solving for the trajectory, the OSIRIS sighting will be much more constraining than navcam. In principle, with some model for the gravitational field, the second bounce location could be estimated by solving Newton's laws. But it's not clear if a linear extrapolation on the orbital images is good, since the lander was following a crudely parabolic path.

About the apparently large change in direction of the lander after the first bounce, most of the motion may have been in the vertical direction and so not obvious from the pictures. In that light the observed direction change may not be very surprizing.

I'd think the way to go rather than assume a parabolic path or whatever would be to use a numerical method with small timesteps and an accurate model of the comet's gravitational field and crank through various scenarios until you get one that matches all the observations.

For changing direction--is it all that mysterious? Even if it came down flat as a three-legged cat, all it would take is an uneven surface under the three feet (like, say, a big chunk under one of them) to send it off in almost any direction, kind of like billiards.

I saw a three legged dog jump out of a (stationary) lorry once!


Veered off to one side, but definetely no second landing site.

Exactly what I had in mind. And with the expertise on this forum, not out of reach. We have good 3D models and total mass estimates. Assuming uniform density as a first approximation, it's straightforward in principle to solve Newton's equations taking the comet's rotation into account. You'd need the initial velocity vector of course, which you could estimate from the stated post-bounce velocity as well as the OSIRIS sighting.

In reality, many of these parameters would be poorly constrained, so you'd end up with a (probably large) uncertainty region around the second bounce location ("landing ellipse")...

Here's a CSV file giving that data every minute:

Philae descent data

The latest SPICE files are from Nov. 15, but they still just have the initial touchdown.

That is great. Here is the relevant info for the OSIRIS image times and the landing time. I guess altitudes are measured relative to some global mean, because the lander "sticks" at 158m upon landing?

15:14 Philae 1337 m, Rosetta 15.2 km
15:19 Philae 1068 m, Rosetta 15.2 km
15:23 Philae 848 m, Rosetta 15.2 km
15:34 Philae 158 m, Rosetta 15.2 km

Those are the numbers that come out--I'm thinking too that the altitudes are relative to the mean surface, so in this case the touchdown point on the actual surface is 158m different in altitude from the reference ellipsoid.

Re: Philae descent data in CSV format.

Does this data represent actual path taken data or is it a predicted trajectory?

Because as per this data the Philae was stationary for over 5 minutes ( 1535 to 1540 UT ) whereas it should have bounced immediately after touchdown.

So that's a roughly 1200 m drop in the 20 minutes to first bounce. Looking at the OSIRIS image, the horizontal motion looks very roughly like 450-500 m over the same interval. So indeed the lander had considerably more vertical than horizontal velocity (at least as viewed from Rosetta).

http://blogs.esa.int/rosetta/2014/11/15/our-landers-asleep/

..."With its batteries depleted and not enough sunlight available to recharge, Philae has fallen into 'idle mode' for a potentially long silence. In this mode, all instruments and most systems on board are shut down."

This implies the lander has active circuitry and some onboard systems have power. What would those be?

Typically spacecraft have some low-level logic (on MER it's known as the battery control board - details here : http://trs-new.jpl.nasa.gov/dspace/bitstre...5/1/01-1682.pdf ) that arbitrates battery voltage, solar array input and deciding to powering on an actual flight computer.

Don't know for sure--it might be the actual path taken to the first touchdown since it's dated Nov. 15 (LORB_DV_055_01_______00098.BSP on this page). It doesn't have any of the bounce data though--hopefully they'll add that when they figure it out.

Does Philae have small radioisotope heating units ~1W (RUH's) like the MER's do to keep the electronics box warm? Also, if the batteries completely drain and then the solar battery recharges some months from now, will the computer be able to reboot so Philae can resume operations?

No radioisotope units on board. See Emily's blog post here http://www.planetary.org/blogs/emily-lakda...-darmstadt.html It contains some information on the potential rebooting of the lander if solar conditions allow.

Turns out the CSV file I posted earlier has errors--I replaced the file with the corrected figures.

Some news from Toulouse's SONC: http://www.ladepeche.fr/article/2014/11/18...c-veillent.html
In French. They say that they know Philae position within 100m but didn't find it yet.

Here are the 6 inserts from the OSIRIS Landing sequence, about 10x with deconvolution

Click to view attachment Click to view attachment

Click to view attachment Click to view attachment

Click to view attachment Click to view attachment

Enjoy

My translation from http://www.ladepeche.fr/article/2014/11/18...c-veillent.html


"We know within 100 meters where Philae is, but we must especially study its orientation, its distance to the orbiter. We'll know more within 15 days, specially with data from the Osiris camera"

"But I remain pessimistic about the sample return. The drill may have dug the vacuum"

I have reconstructed the initial vector that Philae followed after the first touchdown.

Here is a very rudimentary video showing my 3d program and me dragging the timeline.

http://mattias.malmer.nu/wp-content/upload...1/IMG_38991.mov

The blue rays are the projection of philae onto the ground as seen from the orbiter at the times when the images where taken. I reconstructed viewing geometry from the Osiris image using my terrainmodel as surveydata.

The yellow ray is towards the sun from the point where philae cast his shadow.

The intersection blue/yellow sets a point in space where philae was at the time of that picture.

A red ray is shot from the place (in time and space) where philae landed to this intersectionpoint.

This ray then nicely intersects the blue projection of philae as seen by OSIRIS.

Next step is to set up a numerical gravity simulation...

Wow, that looks like a model you can work with.

Just for side interest, I plotted the sub longitude/latitude of Philae and Rosetta during the descent:

Click to view attachment

Interesting to see Rosetta being perturbed as Philae separates--Newton's 3rd law in action. The data points are one minute apart.

According to those results, Philae was traveling just about due south at landing. Annotating an image in Emily's recent blog post:

Click to view attachment

So if this is correct the bounce was to the west.

Very nice simulation, Malmer. If you do gravity simulation, please take into account that the Philae apparently "jumped" roughly in similar direction as the rotation of the comet, so it had more than half of the "orbital speed". So it was more at "suborbital trajectory" than at pure "gravity jump". IMHO centrifugal force has influenced force vectors acting on Philae.

According to your simulation, Philae jumped roughly in the same direction, which flew in and (first) landed on a comet?

Gemini observations of the comet to characterise the activity of the nucleus during landing.
http://gemini.edu/node/12297

Surely that must be the divert manoeuvre that we see in the plot rather that the reaction from separation.

Altitude / velocity / acceleration plots of Philae's altitude component during the first descent, relative to the shape model underlying the SPICE data, based on Joe's revised table:
Click to view attachment
(velocity plot uses average of next 10 minutes, acceleration average of next 10 minutes of averaged velocity, therefore plots are displaced a few minutes)
Using 1-minute deltas would have been rather noisy.

Checking, looks like the Philae mass is 100kg and the total Rosetta mass is about 3000kg, minus whatever fuel has been used, so you're probably right--there wouldn't be much of a recoil from pushing off 100kg.

Any idea about the surface-parallel velocity component after first touch-down?
From the Osiris images my guess would have been about 30 cm/s, corresponding to a 2 km jump in 2 hours.
What's the distance between first and second/third td according to CONSERT?

I will give you my values as soon as i have something i really trust. I want to play with some gravity simulations tonight (CET) if i end up anywhere near the search area I will post whatever my model gives as second landing location here.

The BBC has just reported that the Philae lander has detected organic molecules on the surface of 67P.

BBC News - organics on comet

This looks really nice! I had to play it a few times before I (think I) understood it.

So the left blue ray is the post-first-bounce OSIRIS sighting line-of-sight (LOS) from Rosetta. The right blue ray is the post-bounce navcam LOS. The cyan ball is the lander. The animation starts at bounce, when the red ray actually intersects the bounce location. About 3 seconds in, the right blue and yellow rays intersect the navcam lander and shadow, as expected.

I was baffled for a while why the red ray surface intersection moves after bounce - but that presumably is due to the rotation of the comet? Does this mean that your trial post-bounce trajectory (red ray) is a straight line in an inertial frame? That would probably be a reasonable first approximation, given the weak gravity and short interval (9 minutes) between bounce and the OSIRIS post-bounce sighting. Of course in reality we'd be seeing some curvature due to gravity, and a refinement would be to evolve the red ray as a test particle in the field of the comet.

I'm also very curious about shadows in the OSIRIS frames. As a sanity check, do you project the shadow to be visible in the released OSIRIS image for the post-bounce sighting? What about the three pre-bounce sightings? In that case you'd have to model the pre-bounce trajectory as well.

I have also been playing with the SPICE kernels (in STK) and am trying to create a visualization of the landing. The challenge has been aligning the 3D model of 67P with the correct initial attitude and spin so that the lander ends up at the appropriate spot. Note that what I am doing is a bit convoluted since I am forcing the attitude and spin to match with the SPICE data (as opposed to finding the independently determined properties and hoping they match with the SPICE trajectories).

Here is a quick sample of what I have working so far:

http://youtu.be/AYGcSSlPJqM

The ephemeris data was imported into MATLAB, where I analyzed the final spinning motion of Philae (when it has landed and is spinning with the comet). An FFT yielded a spin rate with period 12.4 hours (matching estimates from other sources, as expected). I crudely estimated the landing position by picking a vertex on the 3D model as close to Agilkia as possible, and from that vertex and the SPICE data I determined the scale needed for the model (I am using the .obj file provided by ESA). The attitude was then determined to align the landing site with the initial position of Philae at touchdown.

This looks nice but is rife with assumptions that could be improved. Here are a list of issues that I am trying to resolve:

- updated SPICE data (there is not a huge difference but it obviously makes sense to use the true telemetry)
- proper 3D model of 67P (more detail is nice, but the real issue is getting the CoG and scale correct... the ESA model origin seems to work but the scale is off by a factor of about 3)
- accurately registered landing position (where exactly is the touchdown position in terms of the comet model coordinate system?)
- accurate comet spin (this should really be defined according to the RA/declination of the spin axis, and I read the spin axis is not surprisingly the principal axis of the largest moment of inertia for 67P)

The interest in this kind of detective work is one of the best things about this forum.

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:

Click to view attachment

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?

IMHO the axis of rotation in the Brian's video corresponds well to the ESA model.

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.



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.

Another video of the visualization, this time tracking Rosetta and then Philae as it lands (at 2048x speed):

http://youtu.be/U-9IHI1lX84

...and then from afar but including the operations timeline at 256x speed:

http://youtu.be/qjI7Oerg48I

Im sorry for the crudeness of my animation. Its me shooting the screen with the iphone while scrubbing the timeline.

You are correct in every point. The red ray appears to be moving, it is due to the comet rotating. It is just a line in space intersecting the two "known" spacetime points and the osiris projection line. It aproximates (in the most crude way) the parabola that an actual gravity simulation would create.

I have started working on a full gravity sim. (I take my shapemodel, voxelize it and let each voxel pull on Philae.) it looks like it could work. Im tweaking parameters. it is the first time I have ever tried anything like this so it is bound to be errors in my thinking.

My first inpressions (horribly early draft) is that philae must have bounced over the crater heading slightly to the south. (South being the dark side of the comet) Passed over the peaks on the far side of the crater as seen from the touchdown location. Landing somewhere on the slope outside the crater almost falling of the head entirely. It is landing in an area where my shape model have no reliable stereo coverage so I can't really say much more. Will tweak more.

That was the impression I got by eyeballing it from the projections posted earlier: that Philae may well have ended up in the area 'over the big ridge' that appears dark on most of the 67p pictures I can find.

Does anyone know of a picture, of reasonable resolution, that shows that area illuminated? Or is that area perhaps so dark that we could rule it out, based the on 1.5 hours of sunlight per comet day reported?

Stevelu asks: Does anyone know of a picture, of reasonable resolution, that shows that area illuminated? Or is that area perhaps so dark that we could rule it out,
based the on 1.5 hours of sunlight per comet day reported?

This picture was posted earlier by CeSinge but I don't have the ESA original.

Philae final resting place in sunlight ?

Given that we know the general landing area, can't the overall terrain orientation of the landing site be derived from the panorama images?
If we know at what time they were taken, the Sun's position is also known.

That would give the overall slope direction and narrow down the possible shadow areas.
Just my 2 cents...

Some info on the lander wake-up mode in the Rosetta Lander User Manual:

it's still unclear whether SD2 drilled on the comet or not. it may just have moved poor Philae
http://blogs.esa.int/rosetta/2014/11/19/di...rill-the-comet/

p.s.: prof Ercoli Finzi, the PI of SD2 was my thesis advisor. and she was kind enough to invite me to attend to the mechanical and vibration tests of the drill in 1999 or 2000...

Has the surface temperature of a comet been measured as it nears the Sun ?, similar to 67P's orbit ?

This sounds promising :

"...As the lander appears to be currently shielded by walls, the local temperature may be lower than it would have been at the chosen landing site. So if Philae wakes up, it might remain operative much longer than expected, possibly until perihelion, which is extremely exciting."

http://blogs.esa.int/rosetta/2014/11/19/di...rill-the-comet/

This is very exciting, Malmer! One question I'd ask is how is the field from each voxel calculated? Do you replace the voxel with a point source, or integrate over a cube? If a point source, you'd need to be careful that the lander is never too close to a point source, which might cause a large deflection when the actual near-uniform voxel distribution wouldn't. This all depends on the voxel size, of course.

And you should be able to check your gravity code by seeing if you can keep Rosetta in a realistic orbit.

Not only was I wrong, but I was completely wrong! However, it was fun to see my theory fall apart under the weight of on-the-scene data.

Another Phil

PHILAE SECOND TOUCHDOWN: GRAVITY SIMULATION #1


http://youtu.be/WF3anN_A1mw


Ok this is stuff done by a non rocket scientist so expect it to be wildly wrong.

I reconstructed the viewing geometry from the pictures from Osiris and Navcam
I reconstructed the sun angle by matching shadows.

I used the touchdown as one known point
I used the intersectionpoint of the navcam projected philae and the shadow projection as a second known point

It seems plausible because using those points make simulated philae fly straight trough the projection of himself from OSIRIS.

I calculated the rotation of the comet body and used the VIRTIS heat image as a source for rotation axis.
I built a gravity simulation using a simple euler integrator.
I used converted my shapemodel into a voxel grid and let each voxel act as a point source for gravity.

I let it all simulate a bunch of times with various parameter changes and here is my best guess for Touchdown #2.