Factorio Forums

I was testing some belt designs for science production and came across an odd behavior. I've got an input belt running underneath my science assemblers via undergrounds, and inserters pickup items from the undergrounds. Today I've found that if these inserters pickup from the underground entrance (not exit), they'd randomly get stuck in some sort of infinite loop, unable to ever pickup any more items.

I say infinite loop, because the inserter is constantly stuttering slightly, as if it's trying to pickup an item, but the item disappeared (to the underground exit) just before the inserter grabbed it. It reminds me a bit of for example Burner inserters trying to pickup from a fullspeed red belt. The belt in question is being fed with a vanilla loader and thus should be fully compressed. Also, the science setup I'm testing uses about 72% of the belt (100% / 1.4), and because all inserters on the belt face the same direction, one half of the belt is fully used and thus flows at max speed. The inserters that get stuck and become unable to pickup items are the first one or two, which will always see a fully saturated, full speed belt lane.

I've included the test save on which this occured (it's a sandbox game). The camera should be centered already on the inserter in question that's stuck. While I tried saving and reloading the game with the inserter stuck, the behavior continued even after reloading, so you should be able to instantly see the problematic behavior. I've also tested the same setup, rotated by 180deg, which had the same problems. A little bit north in the provided map you can see a second red science setup where I've reversed the belt, letting the inserters pickup from the underground *exit* instead. This setup doesn't seem to ever get stuck. This leads me to think that maybe the problem has to do with the different pickup rates for inserters depending on what type of belt configuration they pickup from (see https://wiki.factorio.com/Inserters#Bel ... nserter.29).

What did you do: Have a stack inserter pickup items from an underground belt, with the inserter in front of the underground belt.
What happened: After shortly working properly, he inserter seems to get stuck and can no longer pickup items until the top lane of the belt stops flowing at full speed, or is no longer fully compressed.
What did you expect to happen instead: The inserter to continue to pickup items normally. The stack inserter should easily have enough speed to pickup from a blue belt, and its underground and splitter versions. While there is slight variation in the inserter speed depending on the belt configuration, this shouldn't affect the inserter performance in such a binary way.
How to reproduce: Simply load up the provided save file. More detailed explanation above.
Reproduction rate: always, and persists through save/load.
Screenshot: (can't really show the problem, but does show part of the setup used)

I believe that's a related issue, but not the same. My belt is *fully saturated*, as it's being produced by a vanilla loader. I explained this in my original post.

Looking at the devs' response they deem it not a bug, so I guess that's all for this one.

I was honestly quite pumped to go with a belt base, especially with how many improvements and optimisations were made to belts in recent updates, so I'm surprised at how many inconsistencies and problems still remain.

@Youri: The root cause is the same as the report that Loewchen linked. When picking up from the entrance or exit to an underground belt, the inserter can only grab the item from about half a tile worth of belt. That means anytime the items on the belt are moving quickly, the item will be out of reach before the inserter can react and get into position to make the pickup. If you use a slow belt or the belt is backed up, then the item will be moving slowly enough that the inserter can pick up.

Zavian wrote:@Youri: The root cause is the same as the report that Loewchen linked. When picking up from the entrance or exit to an underground belt, the inserter can only grab the item from about half a tile worth of belt. That means anytime the items on the belt are moving quickly, the item will be out of reach before the inserter can react and get into position to make the pickup. If you use a slow belt or the belt is backed up, then the item will be moving slowly enough that the inserter can pick up.

That leaves me with just one question: why does this behavior differ between the underground entrance and exit? I'd imagine they're both 'half a tile long', but the behavior differs so is the entrance just a couple pixels shorter? What could be the reason for this difference?

Youri wrote:That leaves me with just one question: why does this behavior differ between the underground entrance and exit? I'd imagine they're both 'half a tile long', but the behavior differs so is the entrance just a couple pixels shorter? What could be the reason for this difference?

AFAIK, inserters will try to pick up items on the side they’re entering the tile on. In your tunnel entrance configuration, it means that the inserter will aim for the item furthest away and start extending, but the item will already have disappeared in the middle of the tile before the inserter could reach it. In your tunnel exit configuration, the inserter will aim for the items when they appear in the middle of the tile. The inserter is close to that spot, so it manages to grab the item before it leaves the tile.

Try the same setup with non-moving belts that are fully backed-up and see which items the inserters go for.

Actually I think the logic goes more like this. For the underground entrance, the inserter can see the entire blet tile, but he can only pick up from the first half of that tile. By the time the inserter swings and attempts to make the pickup, the item is now covered by the entrance. For the underground exit, the inserter can see the entire belt tile, but can only pickup from the second half of the tile. By the time the inserter swings, the item has moved far enough that the inserter can now pickup. (Even if it can't pick yet, it knows where the item will be next tick, and can pickup then).

I had this issue today while doing a stream. Occasionally the stack insreters on my low density structure assemblers would just sit around and shake. The workaround fix was just eliminating the underground belts where I could. I happened to be recording the session. Right here