Very nice to see this simulation and interesting how close to "falling off" the head. Thought I'd double check if the known solar declination (about +37 degrees) is being used in this. One can follow up later on and show how the sun angle will change with the cometary seasons?

ESA's landing site estimation via radio (the blue diamond) was posted here http://www.unmannedspaceflight.com/index.p...st&p=215316
I think that the simulation could match it well with minor adjustments, as the probe crosses into "the dark side" at a very low altitude.
A few meters less and it would land in the indicated area.

Wow, cool to see this. Some questions:

How close to the far rim of the large "crater" do your trajectories get? Do you hit the rim for any reasonable (whatever that means) perturbations of parameters?

How long does a typical trajectory last from bounce one to bounce two in your simulations? We know it was about 2 hours.

Does the red ellipse around the landing point represent the uncertainty due to the variations in parameter values that you chose? How did you choose your parameter ranges - in particular, do they match actual uncertainties you've estimated on the parameters? This is important because if your parameter ranges are too small the actual landing ellipse may be much larger.

Someone already posted a rotation axis in RA/Dec - does that agree with what you used? For the sun direction, that's already known from planetarium software/ephemeris - does that agree with what you determined from shadows?

Can you keep Rosetta in an orbit with the right period?

First off. I am not really qualified to do this by any stretch of the imagination. I know images VERY well. But gravity is more of a diffuse thing for me. if you say SPICE i think of cooking... I really should try to get my head around that one day or the other. is there a spicekernels for dummies out there?

anyway I worked a little backwards on this... I basically tweaked gravity until i got a flight time that was right... 1h 50 minutes 54 sec. I do not know the mass of the comet so gravity is just a multiplier of the inverse square of distance to each point at this point. but when i calculated it i got this value for the first touchdown location : 0.00159m/s2 Could it be right? Edit: this value was printed out wrong from my code, disregard it)

The red ellipse is just there to give a sense of uncertanty. I should have made it fuzzier. I have no idea on how to model the actual uncertanties in my model. I would guess that they are huge... but what would be the fun in making a landing ellipse the size of the frekken comet

I hit the rim if gravity is set to high. but then the flighttime is cut in half so that is just not a solution that i even try to use. It is very tight tolerances. If i have lower gravity i end up in space between the two lobes. so it is really only a small sweetspot that makes us hit the comet after the time alotted.

be gentle...

That's nicely done on the gravity sim. The only question I have is does your proposed landing area actually get any sunlight? I can't seem to find any images showing the area and ESA sunlight map they put out when they were scouting for landing sites doesn't show the area as best as I can tell.

Also, how does your value for gravity compare to that generated using the initial mass determination of 1x10^13 kg for the comet?

We must have decent Rosetta orbital parameters out there - orbital period and radius. You could simply tweak the total mass until you produce an orbit that matches those (or solve for the mass analytically from Kepler's law, since from Rosetta's distance, a point source comet distribution is not a bad approximation).

This ignores the influence of the sun, which is probably (hopefully) not a bad approximation, at least for the closer Rosetta orbits.

I have made few diagrams just to see if it is possible to show that Philae can not be on top of the small lobe using only simple liner approximations.

What do we know?

First touchdown 15:34:06 (T+0), zero reference time and point
Post touchdown picture with dust 15:35:32 (T + 86 sec), Location A, 27m
Philae Osiris picture 15:43:51 (T + 585 sec), Location B, 196m
Rim of the small lobe 882m
Second touchdown 17:25 (T + 6654 sec, 1h 50m 54'')
Third touchdown 17:32
Incoming velocity 1 m/s
First bounce velocity 0.38 m/s (Vb)

We actually have two pictures of Philae after the touchdown. The one with raised dust and one black and one white pixel, really is Philae. Touchdown point, and both location are on single line. Picture times are also consistent. We therefore have two independent variables to confirm that white pixel really is Philae. Speeds are almost the same. On location A, only 86 seconds after touchdown, horizontal component of Philae speed is 0.314 m/s (it had no time to change), and on location B average horizontal component of speed up to that point is 0.335 m/s. There is another unused parameter. We have been told that bounce velocity was 0.38 m/s (Philae acceleration sensors). Because we have horizontal component of that speed, it now possible to calculate bounce angle. It is arccos(31/38)=34.28 (40 degrees in pictures is from previous, not so precise measurement, sorry).

What is necessary to stay on the top of the small lobe? We know that up to the point B, 196m had taken 10 minutes (585sec), and we also know that for second touchdown it was necessary almost two hours (1h 50' 54''). But, B/C=0.22. We had spend only 10 minutes on the 22% of distance, speed has not changed between two known points (in fact we have slight increase), and now we should suddenly spend ten times more on the rest 78%. I do not think that such trajectory is possible. Philae has most likely gone over the edge.






I have found only one photograph of the back side when it is not in the dark. Poor Philae

My rough calculations result to the value at about 0.00016 m/s2 (about 10 times smaller then your result, Malmer). My calculation is based only on the known mass (1x10^13 kg) and approximated radius (2 km).

I agree with fredk, that simulation must produce good Rosetta orbital parameters (for example about 8 days of orbital period at 20 km circular distance from center of the comet [or about 0.18 m/s orbital speed at same distance]).

CONGRATULATIONS, Mattias for this incredible work of yours !!!! (and for the diagrams/pictures of Siman)
It's a pity that the best images of the area are still 'sierked', otherwise, you would have found the area of highest probability...

Disregard that value. I made a mistake in my code when i printed it out from there. I will give a better one as soon as I get home to my computer again.

I will try to run a rosetta orbit aswell.

The error that I'm most worried about is the error that comes from having to use an reconstructed osiris image. I rectified the upper corner of the mosaic. but if it was cropped then my viewing reconstruction is probably wrong. And that would throw everything out the window basically...

I'll bring it up again. Is the ESA landing site "diamond" in error?
Could the uncertainty in Z be so large that the diamond actually extends into the back side of the lobe?
Trying to reconcile the numbers and images posted above with the info we have.

Click to view attachment

Thank you for tracking that picture down, depressing though it is. It's exactly what I was asking about earlier.

If your calcs are right, it would certainly explain why it's taking awhile to find Philae with Osiris. I wonder how far down 'the cliff' it would have dropped? Imagine the horizons if Phil ever has the power to hop around and take some more panoramas though.

Afterthought: could the triangulation they were doing during the last two communication passes have been "blind" to such a dramatic location/horizon difference as Philae being <down the cliff> on the <dark side>?

The article from the Sonc said that they know Philae position within 100 m but didn't find it yet (that was 3 days ago).
This could mean that they're waiting for Rosetta to take pictures at the right time (1,5 hour window or so per orbit) and they've not yet get it done because of Rosetta orbit?
Is this make sense?
Another point. The landing pictures shown quite some Philae's details from ~ 20km. I red that Philae was supposed to be 2 pixels but it was a lot more. Could those 2 pixels been it size from 30km?

If Philae is in that notch, it's in the northern "hemisphere" of the comet, right? If so, with the declination of the sun heading to the south in the coming months, the illumination situation may not get much better. Very complicated geometry though to figure out if/when it will get enough sun to wake up.

The OSIRIS narrow-angle camera spec is 20 microradians per pixel. At landing Rosetta was about 15 km away, so that's about 30cm/pixel. I suppose that Rosetta could be commanded to make a close-up reconnaissance run if they know when the lighting conditions are right for a photo.

Rosetta and Philae engineers from ESOC and the Lander Control Centre will do a reddit AMA (Ask me Anything!) today, 20 November 18:00GMT on this subreddit: http://www.reddit.com/r/IAmA/

I don't know how detailed answers they are able to give on such venue but one can always ask, well, anything.

That's about consistent with my first estimates. But we need to be careful with the extrapolation, since due to the height of the jump the rotation of the comet under Philae cannot be neglected. Philae's angular velocity relative to the center of mass of 67P slows down with height. This applies to the velocity over ground, too. For a mean height (integrated over time) of 1 km (as an upper bound estimate) this could be up to about 0.9 km (within 1h50m for a 12.7h rotation period, 1km * 2 pi * 1h50m / 12.7h), which may have to be subtracted, not ruling out, that Malmer's simulations may be roughly correct.
With 0.335 m/s the jump would be 2211 m in 1h50m, minus 900m would be 1311 m over ground.
The value depends mainly on the angle (vertically and horizontally) between the tangent of 67P's rotation and Philae immediately after first td. For a rough estimate needed for the integration a parabolic trajectory should do the job, a better approximation would be a Kepler ellipse. A detailed simulation may return even better results, provided the initial conditions would be known better.

Edit: If the sensors measured the vertical velocity component, we need the arctan instead of arccos, arctan (38/31) = 50.8 degrees.

Edit 2: For the (very much simplified) parabolic trajectory with 50.8° vertical at 1st td, I get a mean height of 452 m ( = 2/3 * (2211m / 2) * 1/2 * 38/31, after some calculus and geometry: mean height = 2/3 max height by subtracting 1/(2a) times integral over x² from -a to a, from a², ascent time = descent time, (mx²)' = 2mx), hence about 405 m to be considered (vectorially) for the horizontal jump distance near the equator. This estimate is most likely too low, since it assumes a homogenious field of gravity.

Yes, I think so. There was a final radio position determination before the probe was out of power. So that would have narrowed down the earlier landing "diamond" to those 100 meters.

Images need to be taken with the landing site illuminated, you can't see the probe in shadow.
Also, some view angles will be blocked by the cliff, and that also limits image opportunities.

https://www.mps.mpg.de/3086295/Philae_Blog1

A couple of short videos roughly showing the illumination on the comet looking down on the spin axis (~5MB AVI files):

Sun at about 45 degrees north declination current declination is 37 degrees
Sun at about 45 degrees south declination very dark situation in the north

set video player on loop to see multiple rotations

Do we know if out gassing can cause alterations of the spin axis and rate of spin as the comet approaches perihelion?

Possibly :

"Evidence is found that the rotation rate of 67P has significantly changed near the time of its 2009 perihelion passage, probably due to sublimation-induced torque."

https://hal.archives-ouvertes.fr/hal-01065970

I took a good look at the images in the DLR Portal for Rosetta Mission's image gallery.
In particular: image #66 and image #75 cover some of the area that was flown over but there is none that shows beyond the crater rim.

Fernando

Sound of first landing as recorded by the CASSE instrument in the lander's feet. Duration 2s.

http://blogs.esa.int/rosetta/2014/11/20/th...d-of-touchdown/

Amazing, true sound at acoustic frequencies.
It sounds like 3 separate 'hits' on the surface which would be consistent with earlier speculation on the impact patterns in the dust, as photographed by Rosetta --
the leading 2 feet hit the ground first, then the body, then the third foot.

Check out this cool 360 panorama projection of the CIVA images I found. Gives you a real sense of area the lander has found its self in.

http://www.360cities.net/image/philae-land...68,-66.37,100.0

A clue from the reddit:

Also this:
I've seen two released and one leaked ROLIS image during the first descent - has anyone seen any of the other four?

One more, in response to whether the lander would be in danger at perihelion:

From reddit, looking for specular reflection to find Philae.

Q:
Even if OSIRIS' resolution isn't enough to directly see Philae now, would it be possible to see sun reflect from its solar panels...?
A:
It's at a distance of about 30km now. As we know when the lander received its sunlight on the single panel which gave it power, then we can plan (and are planning) our OSIRIS images to do a scan in that area at a similar time in the comet day. So yes, you are completely correct. This is certainly one way we will use to identify the lander - the solar panels have a much higher albedo than the background comet.

As best I can tell, Malmers sim seems to land close the equator. Here's a projection of the coordinate system onto a shape model of 67p for reference.

http://planetary.s3.amazonaws.com/assets/i...system_slow.gif

That should bode well for the possibility of future awakening, right? Eventually, the Sun would be directly overhead (as the seasons pass slowly).

If we can estimate where the sun is hiding in the panorama in post #775 and we know the time of the panorama, then we can start to construct a map of how the sun could move in the sky. Once this is done for the landing time, then we can see how it changes with the seasons. This ground view approach can complement looking at shadows with the shape modeling. Any information or guesses for the absolute orientation of Philae? Would it have a similar orientation to what it had at the original touchdown?

From Reddit session earlier today. http://www.reddit.com/r/IAmA/comments/2mw5...ol_and_science/

"First, they said we'd landed on our right side, then on our feet, then on our left side. I've been a firm proponent of the Philae-is-a-cat faction all along. I work with the solar array and can say we have illumination on the lid (facing "up") so we're at least not on our head! According to last info, we may be tilted towards solar array 1."
"I can see shadows on the panels that are within minutes of each other each day (and our sampling rate is just over 2 minutes!). We have panel 2 completely free and see nothing in the CIVA picture, but there are shadows cast on it. On all other sides we see rock faces except underneath where there seems to be a very deep hole! So we were really lucky despite all the bad luck! The calibration on the CIVA cameras isn't exact so I can't give you exact numbers, but the walls look pretty close and they are definitely all higher than the Lander! - VLL

Oh, and we're probably facing NWish with panel 2 since we see the sunset (the sun rises in the East, sets in the West goes over North at noon)"

Excessive quoting removed- MOD

Nice!
Someone ask about your work.
Here's answer;

http://www.reddit.com/r/IAmA/comments/2mw5...science/cm86y7d

I have been very happy to see that several people here have been using SPICE, it opens up some interesting possibilities as this thread has shown (as a different example, it would have been far more difficult for me to process, derotate and stack the Voyager 2 images of Uranus without SPICE).

I would love to see SPICE used more widely here because I think it could result in (even) more advanced things being done by the amateur space image processing community. As far as I know there is no "SPICE for dummies" out there so the idea of writing something like "SPICE for dummies" some day has actually occurred to me - there is a lot of documentation at the PDS NAIF node but it is somewhat confusing, at least at first sight, and a much shorter streamlined guide with emphasis on information I think is particularly useful for the amateur image processing community could be useful. In brief, using SPICE really involves two things:

(1) Knowing about the different kernel types, which kernels to use etc., where to download them and then using some of the programs downloadable at the PDS NAIF node for accessing/using them. The kernels can also be used in e.g. Celestia (or even ISIS for advanced users; requires Linux). This would be sufficient for some users but (2) is what really makes things interesting in my opinion.

(2) Programming against the SPICE libraries that can be downloaded at the PDS NAIF node. This is what makes using SPICE *really* useful, at least in my case, and it's obvious that I'm not the only one. For example, jmknapp has clearly been doing very interesting things here.

Perhaps you've seen this other simulation on YouTube (by flightgear flug), with audio commentary: https://www.youtube.com/watch?v=VeqhQKepzaU
In this one a case is made for a longer orbit down to the side of the larger body. However it lands in darkness (at least at landing time).

This one is updated with consideration of the comet rotation. This maybe goes into a far away area with some small sunlight. Thus does the 110 minute flight time imply a longer arc? And Malmer is right going more off the edge?
https://www.youtube.com/watch?v=M6sRMmzEZyw. This remains inconsistent with the radio fix within the 100m diamond.

Looks like some off the shelf software is being used and he pretty much explains how to replicate this: http://orbitersimulator.com/

Is there a way we can define all of this data in a common reference frame? I've been working with the .obj file and have been wondering how to interpret that 3D data correctly. Here's what we really need:

- What is the best available 3D model for 67P-C/G? Is it the ESA file?

... and then in terms of the 3D model coordinate system:

- Where is the centre of gravity?
- Where is the spin axis? (and how fast is it rotating... early SPICE data suggests 12.4000 hour rotation, but I've seen estimates up to 12.7 hours)
- What are the initial conditions at 1st touchdown? (position and velocity just before the first touchdown)

That would allow us to get an idea of the expected outbound direction since it depends on the surface normal. Of course this is still a crude analysis since details regarding the dynamics of Philae's landing struts interacting with the surface would not be included ...yet

I know that these properties of 67P are out there but it would be nice to summarize them in a common frame so we don't have to reinvent each others' work.

It would also be great if we could do the same with the various photographs people are using to track positions as well as texture the 3D data. Is it possible to define the camera pose and intrinsic properties for each of the relevant images?

A great way to play with SPICE data is to use STK (Satellite Tool Kit) from AGI. You can get it for free and then add "satellites" whose trajectories are defined by SPICE files. The 3D graphics are great for visualization (see my earlier posts), but you can also generate spreadsheets of basically any data you want derived from the SPICE data (ie. position and velocity in various reference frames, orbital elements, etc.).

If anyone does end up using STK, let me know and I will share my scenario files.

The rotation state of 67P/Churyumov-Gerasimenko from approach observations with the OSIRIS cameras on Rosetta
http://sci.esa.int/rosetta/54651-mottola-et-al-2014/

Measuring comet 67P/C-G
http://blogs.esa.int/rosetta/2014/10/03/me...g-comet-67pc-g/

The 12.7 hours is the older value; the rotation period has changed in the last several years, for reasons to be explored.

Unfortunately the spin axis directions disagree between those two references. Maybe someone mixed equatorial with ecliptic.

That paper by Mottola et al. was the first place I looked previously, but unfortunately their estimates are not aligned with the 3D model coordinate system. We would need to know the attitude of the model reference frame at a known epoch in order to determine the spin axis in the model coordinate system. I forgot that they have a more precise rotation speed from that analysis, thanks!

T = 12.4043 ± 0.0007 h

Also, the blog entry "Measuring comet 67P/C-G" does not really give a proper scale and it is difficult to infer the exact scale from the images with dimensions. Maybe one could overlay the image with an orthotropic projection and play around with the orientation until the silhouette matches...

Found a nice diagram of the CIVA camera positions and coverage. It really helps in understanding the panorama we have.
For example, overlap is dependent on surface distance.

Click to view attachment

Using it as a guide, here's a rought panorama. I placed the images by adjusting the (tentative) overlaps to the closest possible FOV intersection points.
The center of the view is related to the camera positions, not the center of Philae.
Rock features and lighting seem consistent.

Here's a 360x180 full panorama.
Click to view attachment

Here's a polar projection, looking down
Click to view attachment

And here looking up
Click to view attachment

To give a clear point of view I am not a friend of leaked data, but I was also mesmerized by this statement. If the data already leaked, and if this statement is published in such a public place, everyone should have a chance to see them. I also tried to find them, with only little success.
If you do a Google image search with: Philae, Rolis, and image size 261x136 pixel you may find one of the images that is probably meant. Don't expect too much, the original ROLIS image once released will probably show much more. It seems to be an analysis of one of the footprints that Philae did on its final landing site.

I am really eager to see the official release of the ROLIS data in original resolution!

I've thought about putting up a SPICE how-to, hopefully geared toward just getting the job done for MER, MSL, etc. but quickly get bogged down in the details. The way I dove into SPICE in the first place was just going through the many "required reading" PDFs at the NAIF web site, which is in itself pretty daunting as it's not clear what progression they should be taken in. The documents are pretty good though, and there is some example code. You have to have a pretty good feel for matrix and vector algebra and coordinate transformations. Then it's a bit of a puzzle to figure out the conventions for a specific mission like Rosetta, what the coordinate frames are, the spacecraft and body names, what specific files are needed and keeping them current. There are a number of programming languages to choose from. It was originally FORTRAN and still used that way by many I think. I prefer C, which may not be everyone's cup of tea or ideal for a how-to. I understand that there may be a java version which might be the best vehicle for a how-to. Brian's tip on STK is intriguing--didn't realize there was a free version of that.

On the subject of languages, I smiled during the reddit when it was revealed that part of the Rosetta code is written in FORTH. I played around briefly with that funky but kind of cool language a long time ago (on an Atari 800). The general word at the time was that it was somehow suited to controlling equipment/instruments and was used at observatories around the world. Job listings for observatories at the back of Science and Nature bore that out. Who knew that has continued into the 21st century, along with FORTRAN, now landing stuff on comets!

BTW, according to the current Rosetta SPICE data, the comet's rotational period is 12.4047 hours. The axis points to RA 69.4160° dec 63.9879° in the J2000 equatorial frame--assuming reality matches the model. Relative to the ecliptic (SPICE frame ECLIPJ2000) the axis points to RA 78.1392°, dec 41.3938°.

.... it was revealed that part of the Rosetta code is written in FORTH ...

That is scary, yet awesome all at the same time! I bet there are still some of those RPN HP Calculators around the ESA facilities also?

Given that in the reddit discussion they said they feel that the lander is still on top of the head somewhere around the crater, I'm wondering if perhaps it did hit the rim and rebounded across the crater.

Check Post #761 for the landing site location.
I mentioned it before, all the official / semi official information gives that spot as the general landing location.

Yes, it's on the "crater" rim and that's consistent with the surface panorama.
Don't see much reason for coming up with alternate possible sites.

I was just discussing what was happening in my silly computer model. I am sure that philae is somewhere in the blue diamond. Right now I'm trying to make my model agree with that. I was very dissappointed when my model sent philae way outside the area. But at the time I could not see why. Now I believe that my initial vector was too shallow and pointing too much to the south. Due to my incomplete idea of where rosetta where at the time.

I'm doing my simulations not in the hope that I will help anyone find the lander but more as a way of trying to learn about these things. It helps to have a little "mission" when trying to learn a new field.

So this weekend I'm going to give it one more try with a better orbiter location...

After that I will revert back to making my navcam animations where I at least know what I am doing...

...and then this:

http://blogs.esa.int/rosetta/2014/11/21/ho...l-landing-site/