Logistic network priority

[bigyihsuan]

This is simply a suggestion for adding an editable priority for logistic networks.

You input a number into 2 text boxes on a logistics chest, one for supplying items and one for storing items, and it will be placed in a list. Anything with the same number will be visited the same number of times by logistic bots.



For example, let's say that we have a robo network its five chests: active provider, storage, passive provider, and 2 requesters. One requester has the priority 10, and the other 5.

The active provider and passive provider have iron in them. As it is now, the active provider will be visited first by robots, then the passive provider.

Next, to drop off the iron, the bot will look for the requester/storage chest that has the higher priority. A requester chest with 10 in the storage text box has a higher priority than a requester with the box at 5, which has more priority than a storage chest, which by default is always placed at the bottom of the list.

So this is a list of priority in the above example, from highest priority to lowest:
Supply (items removed)
1. Active provider
2, Passive provider
3. Storage

Storage (items put in)
1. Requester chest (10)
2. Requester chest (5)
3. Storage chest

Note that the storage chest is at the bottom of both lists.

[ssilk]

Just use two requester chests to double the requests for that facility?

[bigyihsuan]

Just use two requester chests to double the requests for that facility?

But there's not enough space to put 12 requester chests around an assembly machine...

[ssilk]

The logistic bots try to deliver the items randomly, but for all requests evenly. So if you have so few items, that it plays a role to deliver factor 1, 2 or 12 then a prioritized delivery wouldn't change much.

Instead you need to increase the production of the raw materials.

In other words: If you are Coca Cola producer in the mid of summer and you have too less water for your production, then it doesn't makes sense to first deliver Penny market and then Walmart. You need to increase the water production, cause every other decision makes no sense cause if you cannot deliver both, you will loose one market.

[Gandalf]

@ssilk That makes little sense. Of course priorities can be useful, particularly with sparse resources.

Let's say you have a very lacking production of batteries (slow oil pumps or something) which you use for producing laser turrets, accumulators and science packs. Assuming you have an equal amount of chests requesting batteries for each of the three production lines, each will receive a third of the total battery production.
Now you can obviously increase the ratio of one type by adding more requestor chests to it.
But what if you want your laser turret production to always run at full capacity (if at all possible)?
Assuming a setup using priorities, you could assign the requestors of your laser turret production a higher priority then the others. Logistics bots could then bring all batteries to those requesters until their request capacity is met, and only then will distribute any remaining batteries to the other production lines.

Not sure what the point of your coca cola example was but factorio is not an ecomics simulator. Sometimes you want to lose one production line (what you called a market) in favor of a more important one.

After all that I actually don't think that a priorities setup is really needed. It makes sense in theory but in practice I still agree with @ssilk: Increase your input instead of worrying too much about balancing. Adding priorities just means more GUI clutter.

[ssilk]

@Gandalf: replace Coca Cola with green circuits, Penny market with green inserter production and Walmart with red circuits production. You can't really priorize them. If you prefer crafting red circuits, priorize red circuits over the other you suddenly haven't enough green inserters to limit your production. And the other way.

The right way is to limit the production of the end product, and not priorize the delivery.

[y.petremann]

I wanted to suggest this feature but doesn't want to double post.
Actually I think this is a good idea is some point with some scenario, but in my case, I would add the possibility to the storage chest to be filtered (or working as a passive requester with possibility to robot take from it.
I would also add the idea of subnetwork id, so in a network one production would go in priority to its subnetwork but could go elsewhere if overproducing.

[Dev-iL]

cube mentioned to me in private correspondence a few months back that they "plan to do one more category of provider chests with medium priority pretty soon." - so there's that...

The right way is to limit the production of the end product, and not priorize the delivery.

What makes you say that? Does this rely on some theory in the field of Logistics or your personal experience\opinion?

[ssilk]

It's nice to hear, that they plan chests with different priorities... even if it makes things not easier - my opinion.

And what I said is my personal experience with Factorio of course. I think it is clear, that in real life it is eventually different.

[Marconos]

I totally agree with this suggestion. Prioritizing items makes so much sense. I have 3 production lines, First line is for Item X that when I need it I need that item to take priority over other, say ammo for turrets. 2nd line makes Turrets, not as important as making more ammo but when I need them I need them more than other things. 3rd area is standard research production and other things etc. As long as my ammo / turret needs are satisfied then it can run.

Same idea for what chests to empty 1st etc.

Great suggestion and would love to see in the game.

Also, the blanket statement of "increase production" is not valid in many maps and challenge scenarios.

[Plop and run]

One good use case for prioritization is the research. Research is a resource sink that will eat as much as you give it. Also, the more you give it, the more progress you make. What you might want to do is to let your finished goods production take as much as it needs (this production is limited by your construction speed, so the demand is limited as well) and feed the rest to the resource sinks like research. Right now, you have to have separate logistic networks to achieve this and use trains/belts to move resources from mining to goods production to sinks.

Seems like even a simple as a way to separate intersecting (touching) networks will do the job, and this was already suggested.

[ssilk]

Latest with 0.13 you can put the power supply of the labs to a separate network and switch that in/off as you like.

[Marconos]

Latest with 0.13 you can put the power supply of the labs to a separate network and switch that in/off as you like.

Turning lab power on/off doesn't really have much to do with logistics network priority. You are still going to end up with those resources being sent to those requestor chest while the area you deem most important is starving for resources.

[Plop and run]

Exaclty; but even if it did - are you going to shut down your research every time your finished goods factory wants to produce something?

[johanwanderer]

I wanted to suggest this feature but doesn't want to double post.
Actually I think this is a good idea is some point with some scenario, but in my case, I would add the possibility to the storage chest to be filtered (or working as a passive requester with possibility to robot take from it.
I would also add the idea of subnetwork id, so in a network one production would go in priority to its subnetwork but could go elsewhere if overproducing.

A few more requests / ideas:

• Add priority to nearer storage chests.
  For example, I have a large logistic network covering most of the map, and I want to clear a section of the forest to build more. What would be nice is to be able to place a storage chest near the area to be cleared, and have the bots deposit the woods there instead.
  
  Modding idea would be to create a "temporary storage chest" that local construction robots deliver to (and logistic robots may grab from). See below for more details.
• Add a request/storage chest.
  This chest would allow for redistribution of logistic storage. Using the same example above, to expedite the build process in the new area, I would like to be able to place down a request/storage chest for the materials. Logistic robots would then transfer the content from regular storage (or provider) chests to the new area, allowing the construction robots to quickly build a new outpost.
  
  This is something that I *might* be able to mod into the game, if I can modify the robots to prevent them from transporting stuff from a pair of chests.

[bobucles]

You can get FAR better results by setting up your conveyors correctly. When the plates are leaving your smelters, that's where you place all the splitters to set your personal priorities. There are threads that show ways to equally distribute 8 lines, 16 lines, or even get break one line into 3 equal outputs. If logistic bots are handling your most basic resources, well. You dun goofed. Fortunately at that point it's super easy to tear things down and set up anew.

Unfortunately there's no kind of smart conveyor that turns on and off, nor a smart splitter that lets you choose direction. You'll have to wait for future mods I guess.

[sillyfly]

I've been thinking about logistic priority lately, but from a different perspective. Many times you want to have a local buffer handled by logistic bots, say - have a small buffer of repair packs and lasers around certain parts of the walls, so construction bots don't have to travel so much.

Often the though process leads a player to set up a requester chest, with an inserter taking out of it into a passive provider, but of course that won't work - as soon as there are items in the passive provider, the bots will prefer taking from it into the requester, which leads to a sisyphean dance, which may be pretty, but helps no one.

<TL;DR>
I suggest adding a numerical priority parameter to all logistic chests, with a rule that bots may never reduce that number when moving items. So items from chests with priority 1 can go to chests with priority 1,2,3...; but not to chests with priority 0, -1, -2...
</TL;DR>

So this is what I think can be done to solve this problem, which would also allow better control of the logistic network in general (allowing for some form of prioritizing in the original manner discussed in this thread). The basic idea is adding a priority-class to each chest. The priority-class (PC from now on) is simply a number*, with a rule stating that a logistic bot can't move item to a chest with a lower PC. An additional rule for storage chests would be that a bot would consider a storage chest only if all higher PC chests are full.

To help understand this system, consider the following: In the current system, all chests have the same (say - 0) PC (Active Providers may be considered as having negative infinity PC, so bots can always take items from them). If I want to create a local buffer, all I have to do is put down a requester chest with PC=0, and use an inserter to move from it to a passive provider with PC=1 (possibly with red/green wires to only store a certain amount in this specific buffer). Items in the passive provider will never be transported back to the requester, because the rule states bot's can't move items to a lower PC chest. Note though that if there are other requester chests with PC≥1, they may receive items from this buffer, so they are still in the network.

Another use case may be that of prioritizing certain production lines - say I have a limited supply of Advanced Circuits and I can't produce more at the moment. I put a passive provider with PC=1 closer to the output, and another passive provider with PC=0 further down the line. The idea is that the lower priority only receives the ACs if there are enough already in the high-priority chest. The higher-priority factories would be fed with requesters with PC=1, and the lower-priority with requesters with PC=0. An additional measure could be to put requesters with PC=1 leading back into the providers with PC=1, so if there was in the past an abundance of ACs, but now there is a shortage again, the excess (unused) ACs from the low-priority will be moved back to high-priority. This is not required, though, as the bots may take the items from the low-priority providers to the high-priority requesters as is, it just makes sure the low-priority won't get any until there are enough again.
Keep in mind that in such use cases it would be prudent to keep the higher-priority storage relatively small, or else the lower-priority may never get any items at all.

Lastly, the additional rule for storage chests is for prioritizing storage location (as well as the above stated uses). Currently the bots seems to choose the storage chests pretty much at random (some distance seems to be taken into account, but they usually make a mess). This way I can have them deliver first to the highest-priority storage chests, then go down the chain. This may also help visualize how much items I have - if the lower-priority storage is starting to fill up, it means all the higher-priority storage if full.

* I suspect having 3 values (-1, 0, 1) would be enough for all scenarios, but I have not given it too much thought, so better to leave it as a general number.

[starholme]

What do people think about a 'Allow construction bots' checkbox in the Requester chest GUI? So it's just a regular requester chest, but construction bots are allowed to take from it. I understand this only really covers one scenario, but it seems to be a common one.

[edit]As a general solution, I'd love to see that request priority system. Feels a lot like the logistics pipes mod from minecraft.[/edit]

[johanwanderer]

Well, since we don't have a priority system yet, here is my extremely inefficient "bucket brigade" design:

20150710-01-bucket-brigade.png

Basically, the system consists of non-connected logistic networks, connected by a set of request and provider chests. It is very wasteful:

• Everything is doubled because each "cell" can request from two edges and provide to two edges.
• There are at least 4 count of each item in the buffer (at least 1 in each of the requester chests, and 2 on the ground)
• It uses 46 requester chests, 22 on each providing edge and 1 on each requesting edge for back transfer of materials (woods)
• It uses 44 provider, 2 active provider chests, and a few storage chests for good measure.
• It uses 44 pairs of inserters (1 yellow, 1 green) to transfer items, and 2 red inserters to transfer items back.
• The 16 sub-stations/lots of lights/lasers are not necessary, but it lets me plop a solar/accumulator field right on top of it.

But hey, it lets me have:

• Local storage, the bots only travel within their own roboport's range.
• Fast material transfer from the manufacturing base to all outposts.
• Hence, fast construction of outposts.

A few more requests / ideas:

• Add priority to nearer storage chests.
  For example, I have a large logistic network covering most of the map, and I want to clear a section of the forest to build more. What would be nice is to be able to place a storage chest near the area to be cleared, and have the bots deposit the woods there instead.
  
  Modding idea would be to create a "temporary storage chest" that local construction robots deliver to (and logistic robots may grab from). See below for more details.
• Add a request/storage chest.
  This chest would allow for redistribution of logistic storage. Using the same example above, to expedite the build process in the new area, I would like to be able to place down a request/storage chest for the materials. Logistic robots would then transfer the content from regular storage (or provider) chests to the new area, allowing the construction robots to quickly build a new outpost.
  
  This is something that I *might* be able to mod into the game, if I can modify the robots to prevent them from transporting stuff from a pair of chests.

[/list]

[johanwanderer]

I have since switched to a larger, 3x3 roboports (30x30 tiles) cell so I don't waste as much resources on chests and things. Each cell doesn't have to contain all 9 roboports to be effective, just enough to maintain connectivity and pass the materials on to its neighbors. Later on, each cell can become a manufacturing or agricultural base or some such.