1 00:00:11,387 --> 00:00:18,400 MODERATOR: Hello, again, everyone. Welcome back to  the Platypus Hall in PyConline 2021. It is 2021.   2 00:00:20,560 --> 00:00:24,880 Next up we have Mars Buttfield-Addison  with "Python VS Space Junk".   3 00:00:25,520 --> 00:00:33,920 She is a computer engineering PhD student  and has a really cool talk lined up.  4 00:00:33,920 --> 00:00:42,960 >> I am Mars and I thought I would like to say  hello at the start because I won't have time at   5 00:00:42,960 --> 00:00:48,480 the end. My talk goes right up until the end  of the time slot. I will be sitting in chat   6 00:00:48,480 --> 00:00:56,080 and I will take up text questions or via email or  Twitter. I really, really hope you enjoy the talk.   7 00:00:56,080 --> 00:00:58,840 Thank you very much. >>   8 00:01:04,400 --> 00:01:15,360 Hello and welcome to Python VS space junk.  I am a PhD candidate and do a bunch of other   9 00:01:15,360 --> 00:01:23,040 things. My talk is related to my PhD research.  The focus my research is how sensors not made   10 00:01:23,040 --> 00:01:30,720 for space tracking such as made for astronomy  or this can be harnessed to space without   11 00:01:30,720 --> 00:01:36,720 impact meaning it will improve the precision and  reliability of location and collision predictions.   12 00:01:38,160 --> 00:01:43,200 Today I am going to tell you a story about space  junk where it comes from, what we plan to do   13 00:01:43,200 --> 00:01:47,200 about it and what we do about it right now.  There is a little Python at the end that you   14 00:01:47,200 --> 00:01:55,120 can play along on your own time. Let start with  what is space junk and where does it come from.   15 00:01:56,480 --> 00:02:00,000 If you ask any person on the street  most are familiar with about the first   16 00:02:00,000 --> 00:02:11,280 1-5 satellite. You have Sputnik and Vanguard  one the oldest remaining object in space.   17 00:02:12,240 --> 00:02:21,440 They might know fewer more significant ones such  as space stations, ISS, or the hubble telescope   18 00:02:21,440 --> 00:02:28,320 or the hot topic telecommunications such as star  link. Australians of a certain age will remember   19 00:02:28,320 --> 00:02:34,320 when sky lab came down. I am pretty sure we as  a nation are a bit more aware of space stuff   20 00:02:34,320 --> 00:02:39,520 than other nations that don't have a space  program. Those things don't add up to this.   21 00:02:39,520 --> 00:02:47,200 Let's talk about what this is made of and how it  got there. We have launched more things over time.   22 00:02:47,200 --> 00:02:56,240 There are more space capability countries, more  providers, cheaper materials and brought down   23 00:02:56,240 --> 00:03:01,040 the price per weight of getting something to  orbit and more people want a piece of the pie.   24 00:03:03,600 --> 00:03:09,520 We are launching about 1400 objects per year  and about 55-60% of those are just star link   25 00:03:09,520 --> 00:03:20,080 satellites. It is not just more often but we  are launching larger and longer lasting objects   26 00:03:20,080 --> 00:03:24,320 which means they have bigger and longer  laster impacts on the orbit over time.   27 00:03:26,080 --> 00:03:38,640 Space debris doesn't include satellites. Let's put  the beef are constalitons. Debris gets into space   28 00:03:39,520 --> 00:03:46,400 from satellites breaking, launch  waste and astronaut error.   29 00:03:48,000 --> 00:03:53,760 Satellites breaking or running out of power --  there are many satellites that have run out of   30 00:03:53,760 --> 00:04:03,600 power and become inactive mostly after living  a full life and run out of fuel or batteries.   31 00:04:03,600 --> 00:04:09,120 Some fail on launch. They properly detach from the  launch vehicle or battery didn't connect properly   32 00:04:10,800 --> 00:04:16,640 and broken as soon as they left the bay and might  never have fulfilled their experiment or purpose   33 00:04:17,440 --> 00:04:25,600 and now they are debris. We have mass simulators  also otherwise called boilerplate spacecraft. The   34 00:04:25,600 --> 00:04:31,760 tesla roadstar is the most obvious case. It is  going around the sun and might end up in orbit.   35 00:04:35,040 --> 00:04:39,280 A mass simulator is when you need to test your  rocket and need to know before you put people in   36 00:04:39,280 --> 00:04:47,360 the spacecraft that your rocket can handle the  shape and distribution weight. You build a fake   37 00:04:47,360 --> 00:04:54,800 spacecraft that's like one solid plate with  a mass equal to the astronauts you will send   38 00:04:54,800 --> 00:05:03,280 and their equipment. It has to be deployed so I  don't know it will detach and come back down. We   39 00:05:05,920 --> 00:05:12,480 have multi-stage rockets such as the third  stage one here. This one is really funny to me.   40 00:05:13,120 --> 00:05:18,720 A few years ago, there were all these headlines  about how we found this new object. We can see   41 00:05:18,720 --> 00:05:23,280 it. Not very well but there is something in  the earth and we think it is a comet or astroid   42 00:05:24,640 --> 00:05:31,520 because it isn't associated with any launch. It  must be a new tiny captive moon. Oh, my goodness   43 00:05:31,520 --> 00:05:39,840 earth has a tiny moon. When they pointed a high  powered task, it is actually a rocket. None of   44 00:05:39,840 --> 00:05:48,080 our rockets from recent launches are missing. If  we propagate it backwards, this is the stage from   45 00:05:48,080 --> 00:05:56,000 the launch all of the way around and back,  and has been caught on the return flight.   46 00:05:56,960 --> 00:06:02,800 Hilarious. Of course, most astronomers were  like it would be unusual for it to be a comet   47 00:06:02,800 --> 00:06:07,680 or astroid. It was funny to find out it is this  thing that had been on a long space journey.   48 00:06:09,360 --> 00:06:15,520 Also, there is the 100 million tool bag, didn't  really cost 100 million but it was a $100 million   49 00:06:15,520 --> 00:06:24,720 worth of loss where a shuttle pilot was coming to  do repairs outside the ISS and they have special   50 00:06:26,000 --> 00:06:32,240 tool bags and accidently let it go and came down  and ended up in the atmosphere. That was expensive   51 00:06:32,240 --> 00:06:38,320 mistake. But the ISS semi frequently creates  debris because if you have a module you have to   52 00:06:38,320 --> 00:06:43,040 replace and it isn't feasible to bring the old one  back down they have to chuck it off into space.   53 00:06:43,840 --> 00:06:51,600 It is unfortunate but true. Stuff we know where  it comes from gets registered in a catalog   54 00:06:52,720 --> 00:06:55,920 with details of when it was launched  and physical characteristics.   55 00:06:59,600 --> 00:07:07,920 Now a number of debris objects have formed in  orbit. Same amount of mass but broken up into more   56 00:07:07,920 --> 00:07:13,520 fragments. This is dangerous because at orbital  speeds which is thousands of tens of thousands   57 00:07:13,520 --> 00:07:23,360 of kilometers, a break up averages more than 10  items capable of force destruction. Something this   58 00:07:23,360 --> 00:07:31,840 big will create 10 pieces that can smash into  and destroy 10 more things that are this big.   59 00:07:33,200 --> 00:07:36,720 We get this flow effect but  it is a growing flow effect.   60 00:07:39,440 --> 00:07:47,280 In some cases it is exponential on average. In  some pay ways, breaking into smaller objects is   61 00:07:47,280 --> 00:07:51,200 better but they have to be microscopically small  before they stop being an issue on their own.   62 00:07:53,600 --> 00:08:00,560 We have rocket body explosions which are main  debris forming events. We had laws pasted in   63 00:08:00,560 --> 00:08:05,920 decades meaning if you are going to launch  something in orbit with flammable materials   64 00:08:07,920 --> 00:08:12,400 you need to vent it out into space or drive it  until it runs out of fuel because we can't keep   65 00:08:12,400 --> 00:08:17,280 having things blowing up and taking out things  near them and creating enormous clouds of debris.   66 00:08:18,720 --> 00:08:36,240 It didn't stop the explosions. We started calling  them energetic events. Here are examples. This   67 00:08:37,040 --> 00:08:42,880 breeze am in Russian is a good example  because four have exploded since 2007.   68 00:08:43,600 --> 00:08:45,760 The other ones are Chinese and  French repect -- respectively.   69 00:08:47,360 --> 00:08:58,347 Satellites can break apart. A certain design can  be prone to break up. If you launch a fleet they   70 00:08:58,347 --> 00:09:08,080 will break up at the same age. One example is the  transit sats, a precursor to GPS where about half   71 00:09:08,080 --> 00:09:14,800 a dozen have broken up in the same way. Or the  Noah surface and radar sats where I think 15 and   72 00:09:15,600 --> 00:09:20,640 most were recent. They haven't been up there  long but three have broken up into pieces.   73 00:09:21,280 --> 00:09:25,840 Basically big solar panels and lots of  appendages means at low altitude you can   74 00:09:25,840 --> 00:09:29,440 be constantly dragging in the atmosphere  until you fall apart and can't take it.   75 00:09:31,440 --> 00:09:36,880 Now we are up to anti-satellite weapons.  Here are the companies that shot them down   76 00:09:38,080 --> 00:09:43,520 and orange are missile capability countries  meaning you have a missile that can shoot   77 00:09:43,520 --> 00:09:48,240 down other missiles and we theorize would  be capable of taking down a satellite.   78 00:09:53,040 --> 00:09:59,120 In 2008, the U.S. shot down a satellite with a  ship-based missile. The target was a satellite   79 00:09:59,120 --> 00:10:04,560 of their own that had malfunctioned shortly  after launch. The explanation was the orbit was   80 00:10:04,560 --> 00:10:10,640 decaying and it had fuel on board meant it was an  ex explosion risk if it managed to hit the ground.   81 00:10:11,600 --> 00:10:17,360 However, that seemed strange at the time. We  went oh, OK. Even though you were one of the   82 00:10:17,360 --> 00:10:22,320 main countries that made this illegal  you go ahead and break your own rules.   83 00:10:22,320 --> 00:10:28,560 It came out later it was a highly classified  synthetic radar surface surveillance spy that was   84 00:10:28,560 --> 00:10:49,280 the size of a basketball court. The U.S. scored  the first hit in 1985 when a 2B satellite was hit.   85 00:10:49,280 --> 00:10:53,040 There is nothing dangerous about that satellite.  It was just to prove they could, where guess.   86 00:10:54,640 --> 00:11:01,840 Before this -- I guess -- tests from the 1960  said and '70s used thermo nuclear warheads fired   87 00:11:01,840 --> 00:11:09,520 from the ground. That same form factor was later  used to launch satellites into orbit themselves.   88 00:11:11,120 --> 00:11:20,880 Back then they were far less precise but a dead  -- detonation in the atmosphere released a pulse   89 00:11:21,600 --> 00:11:28,080 that would destroy anything within 8 kilo meters.  They want to make sure if the soviets launched   90 00:11:28,080 --> 00:11:38,160 a spy satellite they could kill it before it  hit orbit. In China in 2007 they shot a truck   91 00:11:38,160 --> 00:11:46,000 magnetic kill missile as a test. This event  produced the most space debris out of any   92 00:11:46,000 --> 00:11:54,000 single fragmentation event in history. They made  3,500 plus fragments from a satellite that was 1.5   93 00:11:54,000 --> 00:12:07,240 meters cube thereabouts. India joined in 2019 and  shot down their own and it fragmented less. There   94 00:12:07,240 --> 00:12:15,120 is the classic two things hit each other in space.  The most famous is the showdown in 2009 where both   95 00:12:15,120 --> 00:12:21,360 lost and produced about 2,000 fragments of two  tons of debris. Half of those rocket explosions   96 00:12:21,360 --> 00:12:27,520 I talked about before have already caused full  on collisions with other objects. One took out   97 00:12:27,520 --> 00:12:35,680 another market body and a French one took out a  satellite and the breeze am just a matter of time.   98 00:12:37,840 --> 00:12:40,960 There are also a bunch of near misses  or collisions with small objects.   99 00:12:44,000 --> 00:12:49,360 We only know if one of of the satellites is large  and active enough with cameras on the outside.   100 00:12:51,360 --> 00:12:58,480 Or months later when we figure out the orbit must  have changed and manage to extrapolate and think   101 00:12:58,480 --> 00:13:05,840 it must have been hit with something with this  amount of force. One day they checked out this   102 00:13:08,800 --> 00:13:14,480 and saw the dent and looked at how it happened and  how the orbit changed. Even the ISS gets hit. They   103 00:13:17,440 --> 00:13:25,840 had to dodge debris three times in 2020. They  found something punched through the part that's   104 00:13:25,840 --> 00:13:32,000 responsible for grabbing and docking incoming  spacecraft. They didn't know they had been hit.   105 00:13:33,040 --> 00:13:37,280 Thank goodness the arm is working because  that's an impossible thing to replace.   106 00:13:39,760 --> 00:14:03,200 Here we consider the growth in the number  of objects in orbit from the design grating.   107 00:14:07,440 --> 00:14:18,000 It brings down things already up  there by sending up to already get it.   108 00:14:19,040 --> 00:14:27,680 There are lots of ideas bouncing around. Targeting  with brooms to mostly slow down the object until   109 00:14:27,680 --> 00:14:39,440 the orbit decays. There are more ambitious  graber type ones that look like it is going   110 00:14:39,440 --> 00:14:45,520 up and dragging down the satellites and bringing  them back to the spacecraft. There have been   111 00:14:45,520 --> 00:14:50,960 ideas but none have been tested where they get  something and bring it back to the mothership.   112 00:14:52,480 --> 00:14:58,960 ADR is like a holy grail of space debris research  but no option has proven and is reliable enough   113 00:14:58,960 --> 00:15:11,813 to be deployed. If we keep sending these up and  they are not doing a good enough job and there   114 00:15:11,813 --> 00:15:18,720 is no foreseeable future where we can scoop  it all up because we need things in orbit   115 00:15:18,720 --> 00:15:29,360 despite the risk. We do need a stop gap. What do  we do about it today? We keep track of everything   116 00:15:29,360 --> 00:15:35,360 and use that to predict and maneuver things out  of the way if they go and hit each other. This is   117 00:15:35,360 --> 00:15:41,200 very much like air traffic control but for space.  We call this Space Traffic Management or STM.   118 00:15:42,640 --> 00:15:46,800 The air traffic control is kind of  false because STM is so much harder.   119 00:15:48,080 --> 00:15:52,080 You don't have a transponder on every satellite  telling you what it is and where it is and you   120 00:15:52,080 --> 00:15:58,000 can't use GPS to see where they are. They don't  have reliable tracking of where they were in the   121 00:15:58,000 --> 00:16:04,240 past. Even if we knew where ithorp we can't point  them on the path where they want them to. Most   122 00:16:04,240 --> 00:16:19,280 can't be maneuvered. We can't use a linear path.  You could point them in a straight line and it   123 00:16:19,280 --> 00:16:26,160 doesn't mean they will stick. They are impacted  by solar flairs that change the pressure around   124 00:16:26,160 --> 00:16:35,040 the earth. We put stations all over the globe to  monitor the sky such as this one from Hawaii. This   125 00:16:36,800 --> 00:16:44,320 is a few from all ground base sensors in the  space surveillance network. It is the largest   126 00:16:44,320 --> 00:16:55,120 network of sensors worldwide. Others tend to be  one or a handful of sensors. A few commercial   127 00:16:55,120 --> 00:17:02,560 companies track the stats but sell the data  back to operators for a small fortune. They   128 00:17:02,560 --> 00:17:10,400 tend to concentrate in the northern hemisphere.  Australia can see parts of the sky they can't.   129 00:17:11,840 --> 00:17:19,600 Here in Australia we turned on the first AU based  sensor. A telescope shipped from New Mexico just   130 00:17:19,600 --> 00:17:28,480 for us. They operate a sea band radar from the  opposite side. An Australia company has started   131 00:17:28,480 --> 00:17:37,840 building a dedicated radar array and a few  Australian observing have located space junk.   132 00:17:39,120 --> 00:17:44,800 You are thinking we have so many telescopes  here, why aren't they all tracking space   133 00:17:44,800 --> 00:17:50,000 junk? We have a notable amount. There are  a two simple answers to that question.  134 00:17:52,000 --> 00:17:54,320 Those systems are needed fell --  for astronomers to do their work.   135 00:17:58,880 --> 00:18:06,160 This amounts to why would we waste time  just to do space management tracking poorly?   136 00:18:07,680 --> 00:18:13,360 Some people are working on it. Myself and  engineers are doing experimentation and adaptation   137 00:18:13,360 --> 00:18:21,120 for them and there is a fellow who is working with  physics to do the same. I know people at variousb   138 00:18:21,120 --> 00:18:25,600 universities around Australia are doing the same.  This is a super amount of work and there are key   139 00:18:25,600 --> 00:18:31,600 limitations. One is that the design constraints  and priorities of a telescope designed to look at   140 00:18:31,600 --> 00:18:38,800 space are different than a software and hardware  level. They commonly point at one thing for days   141 00:18:38,800 --> 00:18:46,240 or longer to build up an image that is maybe  only a few pixels wide. The most important thing   142 00:18:46,240 --> 00:18:52,400 is that the telescope can stay as stable as  possible while tracking the earth's rotation and   143 00:18:52,400 --> 00:18:58,400 it can reacquire the object so it can get as much  information as possible out of those few pixels.   144 00:18:59,760 --> 00:19:05,920 STM just wants timely as information as possible.  If you observe this triangle as a representation   145 00:19:05,920 --> 00:19:10,880 of the competing constraints of the imaging  system, we can picture this as the amount   146 00:19:10,880 --> 00:19:18,800 of pixels, colors, and the amount of frames in a  video. More pixels less frames and colors, more   147 00:19:18,800 --> 00:19:29,840 frames less pixels and colors and astronomy is  many pixels and colors as possible, please, space   148 00:19:29,840 --> 00:19:36,000 tracking management is in the give me a snapshot  as often as possible at any cost to quality.   149 00:19:38,160 --> 00:19:46,480 The amounts can prohibit -- the mountains can  prohibit this tracking. Picture my hand as a   150 00:19:46,480 --> 00:19:52,080 telescope. A space tracking telescope could  ideally track something all of the way across   151 00:19:52,080 --> 00:20:00,320 the sky. They can't do that. They have to track  up, and spin all of the way around and go become   152 00:20:00,320 --> 00:20:10,480 down meaning they have a blind sight and that is  not fast enough to track something in the sky.   153 00:20:10,480 --> 00:20:14,560 It turns out even the normal tracking rate isn't  fast enough to hit something in lower earth orbit.   154 00:20:14,560 --> 00:20:22,560 They are just going too fast. If you manage to  track and catch the data you are looking for,   155 00:20:22,560 --> 00:20:27,280 the radio receivers are focused specifically  to eliminate transmissions in the field,   156 00:20:27,280 --> 00:20:31,440 AKA, anything near to earth than the great  distance is it is designed to look at.   157 00:20:34,160 --> 00:20:43,840 An optical telescope can't observe unless it  is in darkness and being lit meaning it has to   158 00:20:43,840 --> 00:20:49,200 be nighttime but near enough to dawn or dusk  so the sun is bethrough -- below the horizon.   159 00:20:54,560 --> 00:20:59,680 If we did a ray trace, from the satellite to the  telescope to the sun, they can see each other   160 00:20:59,680 --> 00:21:04,480 but the telescope and sun can't see each other.  Optical doesn't work because it is too bright or   161 00:21:04,480 --> 00:21:12,320 overcast. A radio telescope needs the reflection  of a transmission that may be preturbed.   162 00:21:15,040 --> 00:21:26,400 Why would an astronomy telescope be built  to rely signals? Very few can. We have this   163 00:21:28,480 --> 00:21:36,960 network responsible for talking to spacecrafts for  NASA whenever the Spain or California stations are   164 00:21:36,960 --> 00:21:42,640 facing their own way. 30% of the time we are  the answering machine for everything out in   165 00:21:42,640 --> 00:21:49,760 space in that direction. Most receivers try to  observe space junk using reflections from these   166 00:21:49,760 --> 00:21:55,760 transmissions but they are so busy wrangling all  the spacecraft their experience is very limited.   167 00:21:57,200 --> 00:22:04,400 Back to thinking OK, fine. We will just build  costume sensors. But they cost billions or more   168 00:22:04,400 --> 00:22:14,640 often billions each and woe need to cover the  globe and have more. They can't be effectively   169 00:22:14,640 --> 00:22:23,360 used for astronomy or near earth imaging for  things such as comets. It is based on doing   170 00:22:23,360 --> 00:22:28,400 that one task and being a good global citizen. You can see how a country might not be the best   171 00:22:28,400 --> 00:22:41,520 at that. How does it work? There is a generalized  process of steps. First you have to identify an   172 00:22:41,520 --> 00:22:45,440 object that you want to look for or a region of  space you want to look at to target a specific   173 00:22:45,440 --> 00:22:50,720 object or objects. Then you have to transmit a  signal if you want to observe a radar response.  174 00:22:51,360 --> 00:22:58,160 Collect data using an optical sensor and analyze  that data and there are many steps of its own.   175 00:23:00,480 --> 00:23:08,240 You need to predict the future trajectory and  adjust the trajectory of objects when necessary.   176 00:23:10,400 --> 00:23:15,200 From step one, you figure out based on previous  locations of objects and predictions of where   177 00:23:15,200 --> 00:23:23,280 they will be, what part of the sky you need  to look at. There is error in cascades. Then   178 00:23:23,280 --> 00:23:34,880 targeting requires conversion between geo centric  meaning if you have a center of the earth acting   179 00:23:34,880 --> 00:23:43,280 as a coordinate system's origin and XY and Z  like a 3d image versus a random point on earth   180 00:23:45,120 --> 00:23:52,480 is the origin and you have coordinates to rotate  north, east or go up altitude. You can see   181 00:23:52,480 --> 00:23:59,840 they don't go one-to-one even if the earth was a  sphere but it is not. It is slightly flat. Also,   182 00:23:59,840 --> 00:24:04,240 the surface of the earth isn't completely regular.  Already that introduces slightly more error.   183 00:24:07,040 --> 00:24:11,840 If you transmit noise, that means finding a ban  on the spectrum you are allowed to transmit on   184 00:24:11,840 --> 00:24:18,000 and giving individual permission for each  transmission that you do to check that this   185 00:24:18,000 --> 00:24:23,200 won't interfere with spacecraft. If big daddy  NASA says yes, you collect the signal in the   186 00:24:23,200 --> 00:24:33,280 big telescope or sensor and that travels a  huge difference back through the atmosphere,   187 00:24:33,280 --> 00:24:38,400 potentially introducing error and the hardware  is prone to error from everything to temperature   188 00:24:38,400 --> 00:24:45,840 to nearby mobile phones to wind speeds to  everything. And digitizing and continuous analog   189 00:24:46,640 --> 00:24:52,080 signal means losing more information. You then  have to use things like Doppler to figure out   190 00:24:52,080 --> 00:24:56,480 where the things are, how fast it is going and in  which direction. You have to get those relative to   191 00:24:56,480 --> 00:25:04,720 earth and doing the translation you did backwards  but also the telescope is on a rotating earth so   192 00:25:04,720 --> 00:25:10,800 it is moving as it is tracking. You have to do  that for every frame and integration window.   193 00:25:12,000 --> 00:25:18,240 You have to figure out what object you point  out and associate it with previous records,   194 00:25:19,600 --> 00:25:24,400 classify how it is moving now because if it  is tumbling or being maneuvered that will   195 00:25:24,400 --> 00:25:31,760 impact the trajectory in the future. Sources are  often precise but the rest can be outright wrong.   196 00:25:33,360 --> 00:25:37,200 You have the prediction of where the satellite  where go in the future which is the hardest part.   197 00:25:40,640 --> 00:25:46,080 The models used to trace the path are complex.  They are called propagation models and have to   198 00:25:46,080 --> 00:25:53,120 take into consideration solar pressure, drags and  earth magnetic fields interacting. Propagation   199 00:25:53,120 --> 00:26:00,720 models -- keep that in mind. Being a complex  physical simulation we don't have the   200 00:26:00,720 --> 00:26:07,120 resources to fully compute to any great level of  granularity. Most use a model called SP4 -- SGP4.   201 00:26:19,360 --> 00:26:22,800 If we want actually useful predictions,  we want to catch it often and   202 00:26:23,520 --> 00:26:29,760 every single one of those thousands of objects.  Say we do that. Every object in space every day   203 00:26:29,760 --> 00:26:33,600 and we have a good idea of where everything  is and will be. Now we move things, right?   204 00:26:35,120 --> 00:26:40,800 Even that's hard. First, you don't get a reliable  pinpoint election -- location of where each object   205 00:26:40,800 --> 00:26:52,160 is. You get a litter -- little error along  the axis. This makes it where the satellite   206 00:26:52,160 --> 00:27:00,800 could be anywhere in here. Jow to check and this  isn't a yes, no but like a weighted probability.   207 00:27:01,360 --> 00:27:08,240 If they do intersect, you have to check if one  or both have the ability to move which doesn't   208 00:27:08,240 --> 00:27:13,040 just mean it was designed with maneuverable  jets but it means does it have enough power   209 00:27:13,040 --> 00:27:20,400 field to do so right now. If yes, compute a new  trajectory and for each potential trajectory you   210 00:27:20,400 --> 00:27:28,720 have to check if it will hit the original thing  or anything down the line. If yes, go back, if no,   211 00:27:28,720 --> 00:27:33,520 you have to calculate the instructions to send  to the spacecraft so it will get on the new orbit   212 00:27:36,480 --> 00:27:41,440 and once you set those instructions you have  to figure out whether it moved into the orbit   213 00:27:42,080 --> 00:27:49,040 or needs further adjustment. If we do all  these things well, we can prevent some   214 00:27:49,040 --> 00:27:53,680 types of collisions between objects where  at least one can move some of the time.   215 00:27:54,240 --> 00:27:59,200 All that work, and that's the best we can do to  keep things safe in orbit. Where does Python come   216 00:27:59,200 --> 00:28:06,720 into this? If we look generally there are three  systems. The things that control the telescope,   217 00:28:07,440 --> 00:28:13,920 maybe a single computer. A thing that does hard  number crunching and this may be a supercomputer.   218 00:28:14,720 --> 00:28:19,680 There is the thing that gets the STM specific  information on the end which might be like the   219 00:28:19,680 --> 00:28:24,480 scientist's computer or the researcher's  computer. Early steps are low level and   220 00:28:24,480 --> 00:28:28,560 need to be super fast and tend to be C-like.  The middle step depend on the other systems.   221 00:28:30,560 --> 00:28:36,560 The last is done by individuals or they are  application scientific like Matt lab or system   222 00:28:36,560 --> 00:28:42,400 tool kits or custom Python. This is recent  happening as new players came on the scene.   223 00:28:43,360 --> 00:28:47,120 Since the '70s and '80s when the only people  before then doing this stuff was norad.   224 00:28:51,120 --> 00:28:55,920 You had devs who had to collaborate with  scientists and physicist. We ended up with   225 00:28:55,920 --> 00:29:02,480 more Python over time. A bunch of scientist  libraries from C, Java or obscure languages got   226 00:29:02,480 --> 00:29:08,080 Python wrappers and we got more dedicated Python  libraries and old tools people made for themselves   227 00:29:08,080 --> 00:29:14,640 started getting appropriate open source licenses.  Here are the most popular nowadays. One is written   228 00:29:14,640 --> 00:29:20,320 by a space organization and the other just by  one person. Commercial companies and military   229 00:29:20,320 --> 00:29:28,240 space tracking organizations and tools such as  these have become so widely used and tested that   230 00:29:28,240 --> 00:29:35,440 people are asking to switch. It is very much like  the comic here except the infrastructure here is   231 00:29:35,440 --> 00:29:40,720 responsible for all of space safety and that one  random person is from Ohio or somewhere in France.   232 00:29:42,080 --> 00:29:46,720 The problem is far from solved because the  limitations of the models are known and new   233 00:29:46,720 --> 00:29:54,000 tracking operators pop up and roll their own.  They don't realize that all of the models suffer   234 00:29:54,000 --> 00:29:58,400 from similar limitations because that's physics.  They can I can roll my on and it will be better   235 00:29:59,520 --> 00:30:06,720 and we can tell -- sell the models. While  we are getting better at data sharing   236 00:30:08,560 --> 00:30:15,920 and using the same underlying libraries for orbit  determination or propagation there is still a   237 00:30:15,920 --> 00:30:21,360 severe lack of standards and nobody contributes  back to core. Even if they are sharing data they   238 00:30:21,360 --> 00:30:28,000 are not sharing implementation because STM data  is worth too much to have people make it for free.   239 00:30:31,520 --> 00:30:36,960 If you own one that can do Space Traffic  Management tests, or you can force a commercial   240 00:30:36,960 --> 00:30:45,840 space traffic provider to contribute back to space  do that. If you know Python you can develop faster   241 00:30:47,040 --> 00:30:52,240 heuristics and benchmark propagation models  and anything new you make against each other.   242 00:30:53,360 --> 00:31:01,520 You can see they introduce different types of  errors. You can develop special case models   243 00:31:01,520 --> 00:31:09,680 for objects that are small or too far away  or have regular orbits or implement easier   244 00:31:10,400 --> 00:31:17,840 GPU and threading support for models because the  observatories are all going to GPU and Cloud.   245 00:31:19,120 --> 00:31:23,760 That will make it easier for the scientists.  If you are a Python dev, lots of these places   246 00:31:23,760 --> 00:31:30,960 need Python devs. The work to get started is  to go with a website with every detection of   247 00:31:30,960 --> 00:31:38,240 every satellite since dawn of the internet in  your preferred format, install the library,   248 00:31:40,640 --> 00:31:45,520 play around and relace different part of the  pipeline. You can take a detection from yesterday   249 00:31:45,520 --> 00:31:50,720 and predict where it will be and pair where it  was seen today and there you go. Space tracking.   250 00:31:52,240 --> 00:31:57,440 If you need hand getting started you can go here.  I have got some code for you. Let me know how you   251 00:31:57,440 --> 00:32:13,840 do. Thank you for listening -- listening. You  can tweet me or email me. Thanks for having me.  252 00:32:13,840 --> 00:32:18,000 MODERATOR: Hello, everyone. Thank you so much  for your talk, Mars. That was really informative.   253 00:32:18,640 --> 00:32:24,800 Mars has been working hard on answering  questions. We are going into a 45 minute break   254 00:32:26,080 --> 00:32:31,840 until the next talk. Get up, stretch your legs  and walk around. I highly recommend people check   255 00:32:31,840 --> 00:32:39,600 out Pycon AU social to see the social events  planned tonight and tomorrow. There is a dinner   256 00:32:40,880 --> 00:32:49,920 meeting tonight and RPG sessions still open for  applications for tomorrow as well as some other   257 00:32:49,920 --> 00:32:55,360 cool stuff such as Mario Cart. Lightning  talk submissions closed and they will be   258 00:32:55,360 --> 00:33:09,840 an at 11:00 a.m. tomorrow if you want to watch.  Check it out. I will see you all another time.