Discussion What is the most efficient way to crowdsource the 3D system coordinates

[Chromatix]

Did you see my code above - I really believe that's the simplest way to do it, just minimise the errors.

It is indeed simple, and it does seem to work well when given good data. The problem is that I can't predict what it'll do if given a single bad data point - will I be able to tell that it is bad, or will it silently give me an approximate position that I can't reliably distinguish from being correct?

Data quality is a perennial problem with crowdsourcing.

The advantage of using algorithms which always give an exact solution is that it is then relatively easy to filter out the data points that don't make sense. I can even check the precision of the result by perturbing the inputs by their range of precision. These sorts of things are interesting to me, even if nobody else thinks so.

With that said, the fact that coordinates are available for all the populated systems suggests that maps would be more immediately useful than coordinate-colletion systems. Making good maps for ED is challenging in its own right, given the differing capabilities of the various ships.

 

[Codec]

The least squares will automatically take into account bad data, although it will make it drift from the true position true, but with enough crowd sourced data this will reduce. E.g., if we get fairly accurate data from 10 systems connections, then this gives even better results than from the minimal 4.

You also get a margin of error from the least squares routines that tell you how well it converged.

However doing it two different ways is good as it cross checks things.

 

[Codec]

Someone has already uploaded some data - so thanks!
The input form is available here

The data is viewable here

 

[RedWizzard]

It is indeed simple, and it does seem to work well when given good data. The problem is that I can't predict what it'll do if given a single bad data point - will I be able to tell that it is bad, or will it silently give me an approximate position that I can't reliably distinguish from being correct?

The least squares method will give you the best fit for all the included data points. This is our estimate of the coordinates for the unknown system. Then what you could do is to take that estimate and recalculate all of the distances back to the reference systems. If the calculated distances match the supplied distances then we know the estimated coordinates are good. If they don't match then find the calculated distance that is most different from its corresponding supplied distance: this data point is suspect. Remove that data point and run the least squares method on the remain points. You should get a better estimate. Repeat the process until you're happy with the result (or you run out of good data).

 

[Chromatix]

I just took a little trip out to Ross 905. I haven't taken a set of distances yet, but it's immediately clear, by comparing it with the map, that the data dump we received is incomplete. It seems to include only "populated" systems, but on the fringes there are a lot of unpopulated ones.

This has two major implications:

1) We *do* need to collect distance information, and turn that into position information, for all these "missing" systems, in order to add them to the map.

2) It is probably possible for even a Sidewinder to reach the other outermost populated systems, even though they are disconnected on the map, by going through unmapped (and unpopulated) systems. However, this will be difficult and risky until the galaxy map starts behaving itself again.

 

[Chromatix]

Well, I can't get to Rahu. It is isolated from all other systems visible in the Galaxy Map by a distance greater than the Sidewinder's jump range. The lack of a coherent route-planner in-game is something of a pain, but at least it's now easier to find out what your ship's actual jump range is.

BTW, I'm seeing reports that the Viper's jump range has been severely nerfed. Can anyone put hard numbers on this before I waste my credits on one?

I did get to the fairly nearby Theta Draconis, and to 41 Gamma Serpentis, and took a set of distance measurements from each. In theory, that gives me enough data to locate the ten external reference points in ED's coordinate system.

 

[Slopey]

I'll have this cut into the BPC early in the next week I'm away meantime.

 

[wolverine2710]

Well, I can't get to Rahu. It is isolated from all other systems visible in the Galaxy Map by a distance greater than the Sidewinder's jump range. The lack of a coherent route-planner in-game is something of a pain.

Might I suggest having a look at "TradeDangerous: power-user trade optimizer" (TD). One of the functions is a route planner. Wanted to have alook at adding the system coordinates but it turns out that Harbinger someone already has done that. His post in the TD thread. You can get more detailed info with the --detail option. So his example with 2x --detail:trade.py --detail --detail nav --ship hauler --full --ly 7.37 eranin wyrd. bash users, see beneat canprefix the commands with time to see how long it takes. His trade route planner is very very fast. example: time trade.py --detail --detail nav --ship hauler --full --ly 7.37 eranin wyrd

Simply install latest python 3.x, optionally install git which gives you gitbash to get a bash shell (unix/linzx) which is far better then a cmd box and install sqlite3 so you can add his new rows to the sqlite database. After one could use his route planner to plan a route to some of the systems Michael Brookes provided. When more coordinates have been calculated they can be added to the database.

Note: Currently on a small vacation with limited internet access but will try to check in from time to time. Probably back at wednesday to fire up my SW.

 

[wolverine2710]

Haven´t been able to follow/understand the thread completely this week but how far are we from taking this to the next level and creating a new thread called something like "Volunteers need for a crowd sourding effort to get the SB2 star coordinates. I can create that thread- if needed - just like a created this one.

I believe Chromatix and Codec have been able to figure out how to calculate the coordinates.
I believe Chromatix has figured out what distances have to be calculated and what the reference points are.
I believe Codec has created a google doc spreadsheet with a form.

What is currently not clear to me. At first there was a discussion that we have to visit the systems with a ship and take distances. Later I read it would suffice that we go to reference points and then use the Galaxy map to find distances to the rest of 570-307=243 star systems. Later I read we have to go out and take distances. What is the status about that.

NEW thread:
The new thread should be real simple and just tell commanders what they should do and what has been done and what is still left. And point ideally them to a webpage where they enter data.

Everonewho has contributed to this thread. THANKS A LOT commanders, really really appreciated.

 

[Snuble]

Getting the space coords when stand at the "unknow" system is doable without resolving to brute force methods.

Its called Trilateration, and in short is getting the intersection from 3 spheres where center is a known set of x,y,z and the radius is the distance to you "unknown" system. The intersection then become 2 points in space (if you got good references, and you will when you stand in the system you want coords for). To get the correct point, test the two results against the 4th known distance, whatever is closest is your answer.

The math is described here, easy if you just did your vector math, hard if its a while since your brain had to think vectors in 3d, or never done math.
http://en.wikipedia.org/wiki/Trilateration

The best implementation example I found was this:
http://stackoverflow.com/questions/23962476/reducing-3d-trilateration-to-2d-trilateration

The methods listed are different vector operations, so more available on the internet.

Tested on the two systems I've visited since Beta 2.03 that where not included in FD coord spreadsheet, and the linked implementation give good answers. Both are not populated.

Note, I've switched z/y and couldnt be bothered fixing, so if I forget to change back please do that.
LP 271-25 -10.46865, 31.84327, 7.31221
Ross 52 -7.46901, 28.19802, 12.78613

 

[fsdraconis]

Chromatix - I can tell you that the Viper starts out with less distance than the Sidewinder when you get it. I don't have the credits to upgrade the engines. The fuel cost for the Viper is also much greater than the Sidewinder.

Stock Viper will start out with 4 ton cargo slot. The engines are nearly already maxed and past the optimal zone. The jump distance with full load of 4 tons maxes out at 6.66 ly distance. It is less than 7 light years. It is over 150k for the first thruster upgrade and have not seen any jump drives to price for the viper.

Got any suggestions for where to find them?

 

[wolverine2710]

Chromatix - I can tell you that the Viper starts out with less distance than the Sidewinder when you get it. I don't have the credits to upgrade the engines. The fuel cost for the Viper is also much greater than the Sidewinder.

Stock Viper will start out with 4 ton cargo slot. The engines are nearly already maxed and past the optimal zone. The jump distance with full load of 4 tons maxes out at 6.66 ly distance. It is less than 7 light years. It is over 150k for the first thruster upgrade and have not seen any jump drives to price for the viper.

Got any suggestions for where to find them?

Thanks for the info.From a thread I got the following response from yanaran: It depends on how much you upgrade. The Viper varies from about 7 when new to about 21 with a D rated frameshift drive (500k credit upgrade). I haven't seen higher versions of the upgrades.

A SW seems a good ship for exploration as well. A snippet from "So why do we need an Asp Explorer? (Sidewinder is awesome!)"":
With a bit of money, and a few mods the Sidewinder can be made into and awesome ship. As of now, I can jump around 15ly, and have a detailed surface scanner. Not to mention with a bit more money I would buy a better jump engine

 

[Mossgarth]

Discrepancies in distance data

I haven't been able to wrap my head around the 'least square' algorithm in 3D space yet, but when experimenting with the linear algebra i remember from school I found some anomalies in the reported distances in #103 when I compared it to the "official" system coords from #89.

LHS 391	Eranin
"official" coords:
LHS 391: (-15.90625, 45.15625, 21.375)
Eranin: (-22.84375, 36.53125, -1.1875)
Calculated distance: 25.13137357
Reported distance: 10

Can anyone confirm? I'm a bit far from that area at the moment.

I also found much smaller differences for (41 Gamma Serpentis, Aulin) (diff 0,27) and (Aulin, i bootis)(diff -3,63).

 

[nicomak]

I haven't been able to wrap my head around the 'least square' algorithm in 3D space yet, but when experimenting with the linear algebra i remember from school I found some anomalies in the reported distances in #103 when I compared it to the "official" system coords from #89.

LHS 391	Eranin
"official" coords:
LHS 391: (-15.90625, 45.15625, 21.375)
Eranin: (-22.84375, 36.53125, -1.1875)
Calculated distance: 25.13137357
Reported distance: 10

Can anyone confirm? I'm a bit far from that area at the moment.

I also found much smaller differences for (41 Gamma Serpentis, Aulin) (diff 0,27) and (Aulin, i bootis)(diff -3,63).

Confirmed by someone from LHS 391 : 25 LY not 10

 

[wolverine2710]

Thanks for the info.From a thread I got the following response from yanaran: It depends on how much you upgrade. The Viper varies from about 7 when new to about 21 with a D rated frameshift drive (500k credit upgrade). I haven't seen higher versions of the upgrades.

A SW seems a good ship for exploration as well. A snippet from "So why do we need an Asp Explorer? (Sidewinder is awesome!)"":
With a bit of money, and a few mods the Sidewinder can be made into and awesome ship. As of now, I can jump around 15ly, and have a detailed surface scanner. Not to mention with a bit more money I would buy a better jump engine

kfsone has just pushed 4.0 to bitbucket where it can be downloaded. It includes the data given by Michael Brookes and work done by others. Might help to find routes to some more systems to use as a starting point. Still on Holiday ;-)

 

[Codec]

Posted a few new system locations worked out from the contributed data.


Input form: https://docs.google.com/forms/d/10YY66ZnpJph8PvyKIFOYcv6amvRHq79SFSybZzQAZE8/viewform

Results: https://docs.google.com/spreadsheets/d/1dNvJV2CnRLU9BqaTpR-lV4D8AW1PfMzOcKzlqIe5aJg/edit?usp=sharing

 

[Berty Dastard]

Don't understand the technicality of what you guys are doing, but respect the effort.

If it helps:

- the hauler with first upgrade (can't remember if it's frameshift drive or engine) will cover 30ly and allows me to go from one end of the galaxy map bubble to the other, in about half a dozen jumps

- I notice that none of the systems prefixed 'wredguia', which crash the game if a visit is attempted, are included in the spreadsheet. Nor are the suffix names (I think)

- also that the number of systems is substantially less than frontier said had been added for beta 2. Although those not visible, inside the bubble and out, are considerable

 

[Codec]

- I notice that none of the systems prefixed 'wredguia', which crash the game if a visit is attempted, are included in the spreadsheet. Nor are the suffix names (I think)

- also that the number of systems is substantially less than frontier said had been added for beta 2. Although those not visible, inside the bubble and out, are considerable

Frontier only gave us data for the systems with stations in, as they had the data handy. The intermediate systems - without stations/outposts we have to work out for ourselves.
You do this by jumping to one of them, then using the galaxy map take distance measurements to 5 or so other systems around, and from that we can work out by a kind of triangulation where the system you are at is in real space.

 

[Trullari]

Btw, can someone edit the results spreadsheet (https://docs.google.com/spreadsheets/d/1dNvJV2CnRLU9BqaTpR-lV4D8AW1PfMzOcKzlqIe5aJg/edit?usp=sharing) ?

I submitted some distances, but it did ignore the decimal mark at first. I resubmitted the correct values, but the wrong values are still there:

http://imgur.com/akPHdMK

This means that the calculated position for LP 180-17 is wrong, sorry.

Fortunately I noticed my mistake after the first system.

 

[Codec]

Btw, can someone edit the results spreadsheet (https://docs.google.com/spreadsheets/d/1dNvJV2CnRLU9BqaTpR-lV4D8AW1PfMzOcKzlqIe5aJg/edit?usp=sharing) ?

I submitted some distances, but it did ignore the decimal mark at first. I resubmitted the correct values, but the wrong values are still there:

http://imgur.com/akPHdMK

This means that the calculated position for LP 180-17 is wrong, sorry.

Fortunately I noticed my mistake after the first system.

No worries - I fixed up the data.