Is that Github link pointing to snippets of source code of software that runs/controls CY2? It was coded in Python? Wow, just wow!chetak wrote:twitter
This zoom-in of the #Chandrayaan2 lander doppler observations show that the start of the fine braking phase seems to have gone according to plan (at 20:18:20 UTC), and something unexpected happened 15 seconds into the fine braking phase.
Source: https://github.com/tammojan/satellite_a ... ding.ipynb …
There is much more to Chandrayaan 2 than just the Vikram lander. It has dual band SAR, a very high res camera and other scientific payloads.Mort Walker wrote:Time to close this thread as images aren’t going to come anytime soon. Vikram is finished and better to move along to failure analysis. Get testing done right whatever the cost. Next big event will have three Vyomnauts go into space for a week and return. The space capsule and all subsystems must be fail proof or people will die.
It will be very slow for information to come out and better to roll this thread into other ISRO discussions.srin wrote:There is much more to Chandrayaan 2 than just the Vikram lander. It has dual band SAR, a very high res camera and other scientific payloads.Mort Walker wrote:Time to close this thread as images aren’t going to come anytime soon. Vikram is finished and better to move along to failure analysis. Get testing done right whatever the cost. Next big event will have three Vyomnauts go into space for a week and return. The space capsule and all subsystems must be fail proof or people will die.
My request is to keep the thread open.
It is simply reading data from some raw files and generating the plots. Python is being used here for visualization and not to control CY2.la.khan wrote:Is that Github link pointing to snippets of source code of software that runs/controls CY2? It was coded in Python? Wow, just wow!chetak wrote:twitter
I was expecting everything to be written in either assembly language or C. I know neitherneerajb wrote: It is simply reading data from some raw files and generating the plots. Python is being used here for visualization and not to control CY2.
...
Coming back to performance, C/C++ beats Python and java hands down. If you can code in assembly then nothing better than that. Extremely fast and compact.
Was excited to see Python as I learnt a little. Thanks for the gyan!execution speed is quite often just one aspect, quite often for rapid prototyping C/C++ ASM are a pain. in debugging....for coding and testing purposes i have seen Python being used more and more often. You never know if people might have used it for prototyping....i know DRDO/ISRO extensively use C++...surprised to see pythonneerajb wrote: Coming back to performance, C/C++ beats Python and java hands down. If you can code in assembly then nothing better than that. Extremely fast and compact.
Wow hearing about PLA after so many decades. Almost forgotten it existance.SSSalvi wrote:After thorough testing & confirmation of logic in C etc, final coding is converted to manually written machine code embedded in Space qualified ROMs/PLAs.
NEW DELHI: US space agency Nasa will try to take images of Vikram lander, lying motionless on Moon’s surface, with the help of its Lunar Reconnaissance Orbiter (LRO) on Tuesday. The US lunarcraft, which has been circling Moon since 2009, will fly over the lander’s site.
The images from LRO will help Indian Space Research Organisation (Isro) know the exact status of Chandrayaan-2’s lander and aid the Indian agency’s efforts to establish contact with it. Vikram had ‘hard-landed’ on Moon’s south pole on September 7 during the 15-minute final descent.
LRO was also close enough during Vikram’s landing attempt to gather data about the manoeuvre using its lyman-alpha mapping project instrument, according to space.com. Reacting to the LRO being used to spot Vikram, the PMO tweeted, “Remembering the ISRO spirit.”
Besides the use of LRO, Nasa has been using its deep space network ground stations to sending hello messages to Vikram lander. “Yes, NASA/JPL is trying to contact Vikram through its deep space network (DSN) as contractually agreed with Isro,” a source in Nasa had confirmed to TOI.
Meanwhile, two foreign astronomers who have been keeping a constant watch on the orbiter claimed that its altitude has come down to around 90km from the 100km circular orbit. Astronomer Edgar Kaiser tweeted, “The left picture (showing a graphic) shows Chandrayaan_2’s 90 km orbit around the current full moon. Note how close to the surface the trajectory is in relation to moon’s diameter. In a week we will have the right picture. Obviously this orbit was chosen to keep the spacecraft in constant sunlight.”
Another astronomer Scott Tilley, who earned fame for finding a Nasa spy satellite that got lost in space in 2005, also tweeted, “As we gear up for observations tonight, it appears Chandrayaan2 may have made a slight orbital adjustment lower its nominal altitude to 92km.” TOI’s repeated attempts to contact the Isro chairman’s office to know about the orbiter’s altitude have failed.
LRO is equipped with a high-resolution camera that it was able to take aerial images of the Apollo landing sites with enough clarity to pinpoint the astronauts’ four-decade-old footprints. The same camera will be used in Tuesday’s attempt to capture Vikram images. The LRO maps Moon and looks for resources that could be valuable for future manned missions there.
Earlier, Isro had taken the help of Nasa’s deep space network to track the journey of Chandrayaan-2 integrated module while it was on its way to Moon from Earth’s elliptical orbit. Nasa’s JPL has DSN ground station in three places —Goldstone, South California (US), Madrid (Spain) and Canberra (Australia). The three stations are located 120 degrees apart on Earth with an aim to ensure that any satellite in deep space is able to communicate with at least one station at all times.
Mort Walker wrote:It will be very slow for information to come out and better to roll this thread into other ISRO discussions.srin wrote:
There is much more to Chandrayaan 2 than just the Vikram lander. It has dual band SAR, a very high res camera and other scientific payloads.
My request is to keep the thread open.
Ravi garu, what do you mean by "support"? Any Turing Complete language "supports" CV. In fact, it can support the computation of any general recursive function. Python is a Turing complete language.Ravi Karumanchiri wrote:^^^^^^^^^^^
Python supports Computer Vision. The entire mission relied on imagery scanned by the orbiter and even by the lander on it's way down. 'Crunching' that in real-time is a feat for Python onleee.
could be ....python is unparallelled when it comes to available libraries..but if its real time i have my doubts..i have not seen many FS using python for real time pricing or other applications..C++/Java rule in that world..hopefully an institutional organization like ISRO should be having a host of libraries for this purpose...Ravi Karumanchiri wrote:^^^^^^^^^^^
Python supports Computer Vision. The entire mission relied on imagery scanned by the orbiter and even by the lander on it's way down. 'Crunching' that in real-time is a feat for Python onleee.
With Python, robotics and other factory automation solutions are more easily paired, when everything is driven by CV and carefully placed 'registration' stickers, the machines can line-up autonomously and pass parts between them, sans computer programmers instructing the robots on how to do it (other than scan for stickers and align orientations accordingly).
Onboard logic cued by 3D imaging referenced to a target map largely generated with fresh imagery while in orbit. Nobody is querying a database on a server farm on earth, all the way from the moon. Python onleeee.
quora doesnt open in office..can anyone post for the lesser mortals...Ravi Karumanchiri wrote: FWIW.....
What programing languages are used at ISRO?
Most of the low-level code happens to be IMKL (Intel Math Kernel Libraries) if Intel processors are used. Other vendors have their LAPACK/BLAS libraries. Most image processing algorithms are linear transformations using small dense matrices. They are embarrassingly parallel.neerajb wrote:... Tensorflow ...
What is the material the balloons are made up of? I bet some very fancy composite metamaterial is required. Remember, Moon doesn't have any atmosphere. So there won't be any reaction force - other than the internal forces of the balloon material itself - to stop the ballons expanding till the membrane gives out.Ravi Karumanchiri wrote:... a number of gas generators would fire and inflate a whole cluster of air bags; as if encasing the lander in a bunch of balloons that would cushion and absorb the initial impact.[/url]
Mars surface is like 33km above Earth. Not much pressure there either. Maybe 0.01 of surface. Balloons fly there quite happily. In fact in the 1950s the Echo 1 and 2 IIRC were mylar balloons placed in LEO.other than the internal forces of the balloon material itself - to stop the ballons expanding till the membrane gives out.
At or approaching the hover stage, there is "pattern matching" to determine if they are at the right place and whether it is safe to put down. That surely appears to be CV, hain? No time for relaying image back and having a Blue Ribbon Commishun with Retired High Court Judges decide etc. The computer decides. Also someone here declared that the entire landing etc is TOTALLY automated, so it must depend on pre-loaded imagery if it is going to compare.ArjunPandit wrote:coming to the computer vision part, is it the the autonomous lander? are you sure ? I would not bet that anything would be CV driven in the landing stage...or am i missing something...i do remember the autonomous part of lander..but would assume it would be sensor driven
i would rate that as a very risky proposition to land an orbiter for the first time using Computer vision, instead of mission controller or other sensors ..its not even that far from earth so that they need to rely on CV...CV has a huge element of predictiveness built in. I doubt ISRO has that data to predict it in those conditions. Even in broad daylight google's moonshot competition for a self drive car coudlnt distinguish between a pebble and a bird at distance (ok its been 5+years)..and here we are talking about far faster speed and in different light,atmospheric (no air etc on moon).
Not funny I know, with all these Serious Anti-Pingreji Experts here racing each other to Report poor me, but the above is precisely what would explain the sudden loss of comms "way up in this cloudbank".Hey! what's Ayesha the Mountain Goat doing, way up in this cloudbank?
The Orbiter/Lander had made multiple passes and scanned the surface. IIRC, ground control had selected a primary and a secondary landing area. The point being they had a good "picture" of the area that the Lander was expected to land at. So, the pixels were loaded into some form of memory - I would imagine.ArjunPandit wrote:coming to the computer vision part, is it the the autonomous lander? are you sure sir? I would not bet that anything would be CV driven in the landing stage...or am i missing something...i do remember the autonomous part of lander..but would assume it would be sensor driven
i would rate that as a very risky proposition to land an orbiter for the first time using Computer vision, instead of mission controller or other sensors ..its not even that far from earth so that they need to rely on CV...CV has a huge element of predictiveness built in. I doubt ISRO has that data to predict it in those conditions. Even in broad daylight google's moonshot competition for a self drive car coudlnt distinguish between a pebble and a bird at distance (ok its been 5+years)..and here we are talking about far faster speed and in different light,atmospheric (no air etc on moon).
this to me seems by some enthusiast if not a silly govt aagint who forgot to delete things..e.g., see wee anal isis of failed iran launch..la.khan wrote:ramana guru this is ot, please delete if need bechetak wrote:twitter
Is that Github link pointing to snippets of source code of software that runs/controls CY2? It was coded in Python? Wow, just wow!
Sun is always pretty shallow at 70.9 deg. lat. on Earth or Moon. So shadows bound to be long. That is the whole basis for ice surviving hajaar^3 saal in craters, hain? But if you are verifying surface features 1 hour ahead of landing, then that was like 8:30AM on a winter din. Very shallow. Which is why I was hoping that by now the sun would have come up as high as it will come up and maybe woken up the solar panels enough to charge the battery and power up the transmitter.ArjunPandit wrote: basic pooch ...why would the sun be shallow? location of landing zone, landing time or the atmospheric effect
also, ironically at the time of landing i was watching 'angel has fallen'
"trained" - that is the operating keyword. I don't think NNs/other machine learning mechanisms are fast enough. My educated guess is that they would not be using any heuristics as weak as NN/ML.ArjunPandit wrote:nraoji, ... which an algo/model trained on insufficient data will reject as match ...
Let us wait for a mech engr./materials scientist to weigh in on this. I don't think this will work for the moon.Ravi Karumanchiri wrote:I realize this animation below concerns Mars, and things would have to work differently for the moon; but 'still and all' this shows the usage of those balloon/bags I mentioned above....
[youtube...]kSbAUtyO7xo[/youtube]
No idea. I can at best guess.ArjunPandit wrote:nraoji, thanks for pointing this out...serious question, wouldnt the resolution and camera be different given the different velocities. Automated operations for such operations would require at least 100s of images for automated verification...slightest perturbations in camera can make the matrix look completely different and result in huge variations, which an algo/model trained on insufficient data will reject as match and will take other alternative or corrective actions...i agree they might have other images..but there would be too many variables at play. Pardon my ignorance, if i am asking very stupid questions
....................
I wonder if Vikram had similar or the same logic.How Terrain-Relative Navigation Works
* Orbiters create a map of the landing site, including known hazards.
* The rover stores this map in its computer "brain."
* Descending on its parachute, the rover takes pictures of the fast approaching surface.
* To figure out where it's headed, the rover quickly compares the landmarks it "sees" in the images to its onboard map.
* If it's heading toward dangerous ground up to about 985 feet (300 meters) in diameter (about the size of two professional baseball fields side by side), the rover can change direction and divert itself toward safer ground.
Why Terrain-Relative Navigation is Important
Terrain-Relative Navigation is critical for Mars exploration. Some of the most interesting places to explore lie in tricky terrain. These places have special rocks and soils that might preserve signs of past microbial life on Mars!
Until now, many of these potential landing sites have been off-limits. The risks of landing in challenging terrain were much too great. For past Mars missions, 99% of the potential landing area (the landing ellipse) had to be free of hazardous slopes and rocks to help ensure a safe landing. Using terrain relative navigation, the Mars 2020 rover can land in more - and more interesting! - landing sites with far less risk.
How Terrain-Relative Navigation Improves Entry, Descent, & Landing
Terrain-Relative Navigation significantly improves estimates of the rover's position relative to the ground. Improvements in accuracy have a lot to do with when the estimates are made.
In prior missions, the spacecraft carrying the rover estimated its location relative to the ground before entering the Martian atmosphere, as well as during entry, based on an initial guess from radiometric data provided through the Deep Space Network. That technique had an estimation error prior to EDL of about 0.6 - 1.2 miles (about 1-2 kilometers), which grows to about (2 - 3 kilometers) during entry.
Using Terrain-Relative Navigation, the Mars 2020 rover will estimates its location while descending through the Martian atmosphere on its parachute. That allows the rover to determine its position relative to the ground with an accuracy of about 200 feet (60 meters) or less.
It takes two things to reduce the risks of entry, descent, and landing: accurately knowing where the rover is headed and an ability to divert to a safer place when headed toward tricky terrain.
Very sudden end with no warning.Controlled Flight Into Terrain.
thanks for the pointer. some folks dismissed these ideas as "flights of fancy". it all depends on the initial boundary value problem and the material properties of the material the balloon complex is made up of.rahulm wrote:Russian Luna 9 in 1966 was the first ever probe to soft land on the moon, it used landing bags