Large logistic networks in 2.0

[Tinyboss]

I've been playing 1.x for so long that I've developed the habit of using bots as a last resort, just to inject some challenge back into things. I decided to let go of all my preconceived notions, as much as possible, for SA and just go at it like a new player taking the easiest and fastest path through. I can do the self-imposed challenge thing later!

So I just covered my Nauvis base with a single logistic network to see how well it works now, with the bot improvements. Still not great, it seems...often a bot flies all the way across the base to do something like construct a power pole. So my question is, what are the new guidelines for an efficient network? Same as before, where we keep each one small and use trains or circuit logic to pass resources between them? Or is it supposed to work better now, and I'm doing something wrong?

[mmmPI]

Same as before for the guidelines, that's what work best.

[shopt]

I always read those improvements as making large bot networks "less bad" rather than good. The biggest change is that you can use the new roboport logistics request to avoid bot starvation in important locations. And I'm now less strict about avoiding all concavity in my network. But I don't think the overall guidance to have separate networks bridged by trains or belts has substantially changed.

[MeduSalem]

Yepp, same meta with separate smaller networks like before SA.

I mean, in the end if you build a huge network spanning the whole map you kinda have to expect that bots over time will eventually start accumulating in a certain high traffic area. And then if you do something on the completely other end of the network, obviously it will take them time to cross the map then. ^^

There is pretty much nothing one could programming-wise do about this. The devs cannot expect in advance how you will build your network and where you want the bots to gather. So you cannot avoid that logistic bots will gather in one particular spot of the map.

I mean they could program it such that some bots will try to spread out over all available roboports so that at least some are in every single roboport so it might pick the closest one for a task, but that seems mighty annoying in high traffic scenarios when all of a sudden bots come in from all over the map because the demand is so high.


So in that regard, I suggest for high-throughput areas to make them their own separate little network where the bots cannot leave. And connect them to the outer world via trains, belts... or logistic chest-to-chest where a request chest is in one network and the other chest a provider chest in the other network and a stack inserter moving stuff in between with barely 1 tile gap between both networks.

And for the base expansion outward into the wilderness have its own network that follows along the train routes (that is what I do and how I expand the network). That one can be huge because it doesn't matter. Nothing will be transported by logi bots there or at least only small amounts maybe. That area is rather meant for construction bots only and obviously those will also gather in the spot where you last built something.

[Tinyboss]

Thanks for the replies!

My problem isn't bots accumulating in one part of the network. Rather that bots/chests are chosen in very suboptimal ways. For example, items may be taken from a purple chest and stored in a yellow chest at the far opposite side of the base, even though there is plenty of available storage nearby. Likewise, when I plop a power pole, a bot on the other end of the base may grab one and bring it over, when I know there are bots and power poles available nearby.

[qoalabear]

The behavior of storage chests is working against you here. From the wiki https://wiki.factorio.com/Storage_chest

Logistic robots will pick up items in the following priority: active provider chests > storage chests, buffer chests > passive provider chests

Logistic robots, and construction robots fulfilling deconstruction orders, will fill storage chests in the following priority:
Storage chests with a matching filter
Storage chests with a matching inventory
Storage chests with no filter and no inventory
Storage chests with no filter and a mismatched inventory

So robots are taking things across the base to far off storage because that's where the existing chests with that item are. You might be able to accommodate this by setting filtered storage chests anywhere that actively provides something but should still be kept locally (though if it fills, you'll go back to the previous behavior). I haven't personally tried to do this, so your mileage may vary.

And bots are probably sourcing materials from the far off chests because storage is higher priority than passive provider for pickup. Depending on what you're doing, buffer chests might help with this.

[Tinyboss]

Good info, thank you! I took a couple years off prior to 2.0/SA and I forgot all the gory details.

So maybe I could make a blueprint with a set of buffer chests that request all the common items, and place these around my base? Then when I construct a power pole, it would be brought from the nearest one?

It wouldn't fix the problem of a bot coming from across the base to do it...or would it? If the chest is chosen first, then a nearby bot is chosen, that would work I think.

[qoalabear]

So maybe I could make a blueprint with a set of buffer chests that request all the common items, and place these around my base? Then when I construct a power pole, it would be brought from the nearest one?

With the new logistic groups (towards the end of FF#382) (and assuming your set of "common items" is few enough stacks to fit in one chest)
You could make a "Basic Base Materials" group, so that you can update all such buffer chests at once! You can also pick low values for the group and have the chest request a multiple of that group (maybe a very far off location wants to hold 10x or more of the standard request to account for the long resupply time).

It should dispatch the nearest bot to the start of the task (in this case, picking up construction materials). You can set a request in the roboport for a minimum number of bots to be stationed in it. Maybe distribute such hubs in the same blueprint as the buffer chest.

[r3xk3nnw4y]

The behavior of storage chests is working against you here. From the wiki https://wiki.factorio.com/Storage_chest

Logistic robots will pick up items in the following priority: active provider chests > storage chests, buffer chests > passive provider chests

Logistic robots, and construction robots fulfilling deconstruction orders, will fill storage chests in the following priority:
Storage chests with a matching filter
Storage chests with a matching inventory
Storage chests with no filter and no inventory
Storage chests with no filter and a mismatched inventory

So robots are taking things across the base to far off storage because that's where the existing chests with that item are. You might be able to accommodate this by setting filtered storage chests anywhere that actively provides something but should still be kept locally (though if it fills, you'll go back to the previous behavior). I haven't personally tried to do this, so your mileage may vary.

And bots are probably sourcing materials from the far off chests because storage is higher priority than passive provider for pickup. Depending on what you're doing, buffer chests might help with this.

While that's good to know, I think the logistics network would greatly benefit from priority drop off with distance. If an active provider 200 tiles away has the power poles you asked for and a passive provider chest 20 tiles away also has power poles it would be a lot nicer if your bots pulled from the passive provider chest rather than flying 200 tiles out for the active provider. Because currently it seems more like you are discouraged from using buffer chests if you have active providers in the same network.

[NineNine]

While that's good to know, I think the logistics network would greatly benefit from priority drop off with distance. If an active provider 200 tiles away has the power poles you asked for and a passive provider chest 20 tiles away also has power poles it would be a lot nicer if your bots pulled from the passive provider chest rather than flying 200 tiles out for the active provider. Because currently it seems more like you are discouraged from using buffer chests if you have active providers in the same network.

That's a discussion as old as time itself (or Factorio). The reasons that bots fly the way they do are based on lots of different factors, and most of them are performance-based. You can read some of the nitty gritty details in this discussion: viewtopic.php?f=71&t=49429&p=314306