Hello.
I have an idea for logistic network improvement. Let say we have huge (very huge) factory and random choising of receivers-providers in logistic network could be problematic...
... first of all, we have to realise the problem. If we always choice a nearest source (instead of random) to manage request, there must be managed flow between sources if some source is burden more then others which consumes another logistic potential. In that meaning, the random choising of source-request is not so bad in comparison with nearest choising in long time behavior.
But there is still problem with many sources and many consumers where unlogical ways are chosen
Description:
Let have some value X related to each "storehouse" in our very huge logistic network. Each storehouse has it own value X:X = CONTENT - REQUEST
Note that REQUEST its only internal value of each storehouse that is not sended into logistic network! Now lets calculate logistic request and logistic content that is sended into network:
LOGISTIC REQUEST = Xn/N - X
Where Xn is sum of X values of its nearly neighbours and N is a number of neighbours (its wise to set 6 neighbours and use hexagonal distribution of storehouses). All logistic request is something that is actually send out and if the value is negtive, our storehouse will provide this amount of items. It could be realized by juts little smart chests setup.
How it works:
It is actully little bit hard to explain: It has two behaviours: Short time where we observe how the fast changes in our network is realized. And long time where we observe how content of all storehouses changes. I belive that our method improves both of behaviors. It is very important to realise that X value is not request nor provision but only internal value of each storehouse. The real logistic request is represented via formula upward (Xn/N - X). The real logistic request is actually zero in all storehouses till we remains in equilibrium state. Which means literally most of the time. If we change X value slightly (by consuming some amount of source), the signal will be spread tick by tick over all network. The signal will reduce the logistic requsest of all neigbours one by one and will trigger at first place where we obtain negative logistic request. It is the first neighbour in general, however there is some cutoff in dividing and this is the most esential fact in whole method. The storehouse will not trigger imidiately but when the difference is high enough. So if we miss 6 items, the 6 nearest neigbours take -1 value into its logistic request and provide items into center. Well, if we meet the same situation again, the nearest storehouse will not be triggered again becouse the value of neigbours content is lesser. It passes far into first storehouse that is able to provide some content. This should couses that every change of request (or contant raising) is spread by tick-by-storehouse speed over all network and trigger in nearest possible source.Lets formulae an "equilibrium state" more precisely. This is the state where all consumption and generation is in balance. In this case, our system should work perfectly: it should assign nearest source to each consumer and only if the consumption is higher then generation, the consumer will find another source and split its requirement into them. Rest of the storehouse will remain passive. If we left the equilibrium state which means there is some change in generation or consumption, the whole network will redistribute contents between storehouses but not so heavy. Its the same like in the temeperture analogy: we open the window and observe how the temperature distribution changes. So works of all the network could be separated into two main process. First of it is distributing materials between sources and consumers in ideal way and the second one is draw/raise of all storehouse content where consumption or generation is higher.
Problem with implementation:
This idea has a little bit problem for me. Its designed for huge factory but i have a problem that my largest factory has few chunks only
. So before i start to build huge factory and start to test this kind of method, i'd like to start dicuss with someone who is interested. There is probably another ways of otpimising.