Clusters chokepoints

Describe your experience with the latest version of FreeOrion to help us improve it.

Moderator: Oberlus

Forum rules
Always mention the exact version of FreeOrion you are testing.

When reporting an issue regarding the AI, if possible provide the relevant AI log file and a save game file that demonstrates the issue.
Post Reply
Message
Author
User avatar
MatGB
Creative Contributor
Posts: 3310
Joined: Fri Jun 28, 2013 11:45 pm

Clusters chokepoints

#1 Post by MatGB »

Two pics:
clusters-chokepoint-1.png
clusters-chokepoint-1.png (197.97 KiB) Viewed 716 times
clusters-chokepoint-2.png
clusters-chokepoint-2.png (274.31 KiB) Viewed 716 times
As I hope is clear in pic one, the entire north of the map is separated from the south by one set of lanes with a couple fairly long jumps. I've now wrapped up and conquered all the northern AIs, so every single one of my ships is going to have to fly through that small number of systems.

While I'm quite happy that this happens sometimes in a random setup, it does seem, or I perceive it, that this happens in some way slightly too often if I'm playing Clusters, and while on a small number of systems map it's acceptable, on 400 systems?

This is just a general feedback thing, not a major issue but I thought better document it in case there's an easy fix and/or it's considered a bug by others.
Mat Bowles

Any code or patches in anything posted here is released under the CC and GPL licences in use for the FO project.

User avatar
Geoff the Medio
Programming, Design, Admin
Posts: 13587
Joined: Wed Oct 08, 2003 1:33 am
Location: Munich

Re: Clusters chokepoints

#2 Post by Geoff the Medio »

Is the red empire entirely cut off from the rest of the galaxy? (Edit: nevermind, I see some systems at the bottom left aren't explored yet...)

Were there any experimentors present at a location that could have removed some otherwise important connecting lanes?

User avatar
MatGB
Creative Contributor
Posts: 3310
Joined: Fri Jun 28, 2013 11:45 pm

Re: Clusters chokepoints

#3 Post by MatGB »

Not that I've seen, no, I've got Omni Scanner and have scouted the edges of all my space. I don't want to do Super Testers yet, the game's still fun.

And a turn later there's a patch of blue on the starlane leading out of Red's territory into the unexplored bit, so it does meet up with the area I've explored, it's just the one chokepoint from Jinga to Scarab that bothers me.
Mat Bowles

Any code or patches in anything posted here is released under the CC and GPL licences in use for the FO project.

User avatar
Dilvish
AI Lead and Programmer Emeritus
Posts: 4768
Joined: Sat Sep 22, 2012 6:25 pm

Re: Clusters chokepoints

#4 Post by Dilvish »

Geoff the Medio wrote:Is the red empire entirely cut off from the rest of the galaxy? (Edit: nevermind, I see some systems at the bottom left aren't explored yet...)

Were there any experimentors present at a location that could have removed some otherwise important connecting lanes?
I don't think the red empire's connection is really the issue he's concerned with, or necessarily an issue at all-- Mat simply hasn't gotten visibility of the full galaxy in these screenshots, but you can still see that the red empire is connected to the light blue empire, which in turn (on the other side of a big blank area Mat hasn't seen yet) is connected to the pink and dark blue empires in the bottom center.

If I understand you correctly Mat, you're just concerned about the high number of super-tight choke-points, essentially, nearly all the clusters are connected by tight choke-points. The AI doesn't yet really explicitly plan for them, but Morlic has come up with a promising idea that we'll be developing more after 0.4.5 goes out. So I don't think this characteristic of the Cluster galaxy shape is really a problem, it's more the defining characteristic. If you managed to make the connecting corridors shorter, then *maybe* that would be better, I'd have to see it & try it out.

But I suggest giving us a bit more time to get the AI dealing with choke points, and then revisit the issue.
If I provided any code, scripts or other content here, it's released under GPL 2.0 and CC-BY-SA 3.0

User avatar
Geoff the Medio
Programming, Design, Admin
Posts: 13587
Joined: Wed Oct 08, 2003 1:33 am
Location: Munich

Re: Clusters chokepoints

#5 Post by Geoff the Medio »

How does the same galaxy look if generated with different starlane frequencies?

It's possible that the gaps between clusters are just too long to have any lanes; currently the value in universe_tables.txt is 120 uu.

During galaxy generation, during the starlane creation step, it loops over lanes, starting with the longest, removing lanes that are longer than the max in that table, when doing so doesn't make to disconnected subgraphs. This is prone to leaving a single choke point in galaxies where a lot of the potential lanes are quite long, such as is often the case in cluster galaxies.

Increasing the value in the table can remove a lot of the single-choke arrangements, but then it's not much of a cluster layout anymore...

It could also be modified to work like the later code that removes lanes only if doing so doesn't make the systems on either end too far apart. But that would probably also prevent many of the galaxy shapes from looking how they're supposed to. Removing the longer lanes is what makes them look like what they are.

User avatar
MatGB
Creative Contributor
Posts: 3310
Joined: Fri Jun 28, 2013 11:45 pm

Re: Clusters chokepoints

#6 Post by MatGB »

Geoff the Medio wrote:How does the same galaxy look if generated with different starlane frequencies?

It's possible that the gaps between clusters are just too long to have any lanes; currently the value in universe_tables.txt is 120 uu.
Haven't tried that yet, but I have used Dilvish's range circle set at 120 (a number I knew although was unsure on how it worked) to go around the edges, and lots of closest points are about 130 to 150 apart.
During galaxy generation, during the starlane creation step, it loops over lanes, starting with the longest, removing lanes that are longer than the max in that table, when doing so doesn't make to disconnected subgraphs. This is prone to leaving a single choke point in galaxies where a lot of the potential lanes are quite long, such as is often the case in cluster galaxies.

Increasing the value in the table can remove a lot of the single-choke arrangements, but then it's not much of a cluster layout anymore...
You'd still get the stars in clusters with gaps between them.

Like I said and Dilvish says, my biggest concern is that having got the whole of the North I now have to fly all my ships through one starlane to get to the rest of the map, I have no flanks and am effectively secure, I have basically 'won' at this point as there's no way any AI will be able to threaten my control of that single chokepoint even if they tried as the sheer weight of all the ships I ahve (let alone the ships I'm building with my PP) is going to be too much.

While I know the AI doesn't handle chokepoints well yet (and have been following some of the discussion about planned improvements), I also think this is potentially a problem in the generation of the galaxy itself, even without Experimentors it creates too much of a cutoff, and with them they could potentially have setup on Scarab Alpha and divided the map in two.
Mat Bowles

Any code or patches in anything posted here is released under the CC and GPL licences in use for the FO project.

User avatar
Dilvish
AI Lead and Programmer Emeritus
Posts: 4768
Joined: Sat Sep 22, 2012 6:25 pm

Re: Clusters chokepoints

#7 Post by Dilvish »

I'll repeat though, I think this is probably fine as a characteristic for this galaxy shape, and I absolutely definitely don't think it's worth putting more time into right now.
If I provided any code, scripts or other content here, it's released under GPL 2.0 and CC-BY-SA 3.0

User avatar
Geoff the Medio
Programming, Design, Admin
Posts: 13587
Joined: Wed Oct 08, 2003 1:33 am
Location: Munich

Re: Clusters chokepoints

#8 Post by Geoff the Medio »

If this is a problem, it's something that wormholes could help resolve. Also giving players the ability to create (and remove) starlanes.

User avatar
MatGB
Creative Contributor
Posts: 3310
Joined: Fri Jun 28, 2013 11:45 pm

Re: Clusters chokepoints

#9 Post by MatGB »

Bringing this up again.
clusters-zigzag-seed-5.png
clusters-zigzag-seed-5.png (290.41 KiB) Viewed 578 times
This is turn three with me just running the supertester takeover. I've completed this map through, and basically had to use Bores and a Nexus to not die of boredom. If it's not obvious why from this pic (I started bottom left), then the following should be clearer (fear my mighty art skills):
clusters-zigzag-seed-5-inked.png
clusters-zigzag-seed-5-inked.png (325.11 KiB) Viewed 578 times
The red areas are impassable, the green lines follow my fleet movement. Basically, once red is defeated and I'm moving into light blue, all my fleets are in the same place going in the same direction. I have no flank and no rear, I'm completely safe.

The steamroller was massive and every time I absorb an AI I jump ahead in tech and thus become even more unstoppable. I opened up extra starlanes simply because the boredom was too much for me.

I like the idea of Clusters as a layout, but it really needs to ensure that each cluster except at the extreme corners has at least two and preferably three entrance/exit routes so you've got at least one flank and you don't get a single, large S shape to fight along.

This isn't a high priority problem, but I've played several different seeds on High starlanes now and it's fairly similar in all of them, I think/hope that when the generation scripts are moved entirely to Python then Clusters can be adjusted to prevent this from happening as often and as badly.
Mat Bowles

Any code or patches in anything posted here is released under the CC and GPL licences in use for the FO project.

Post Reply