experiment with ac loops

[michielbdejong]

• [x] generate network and loops
• [x] send current through loop until numPosBalances-numNegBalances <= 1
• [ ] display results

[michielbdejong]

making this distributed:

• loop finding should use probes
• as a node, if you have an interest in adjusting the balance of a loop, propose it, with a min and a max change. this proposal gets passed around until it gets back. the min becomes a dc current and the diff between min and max becomes the amplitude of an imaginary ac current.
• if you have two loops going through you in a Y, and the stem of the Y is at its limit in either direction for one of the loops, the other loop can help! (i should draw an example of that)

[michielbdejong]

research question: we know nodes can collaboratively find the global optimum. but how can they cooperate using pricing? you would need to think of a way to avoid staring contest extortion, where one party puts their foot down and refuses to cooperate until the other party agrees to a higher price.

[michielbdejong]

This turned out to become slow quite quickly for larger networks. The more promising approach would be netting at equilibrium, where each link has a "voltage" at each side, and nodes can adjust these voltages to incentivize money flowing through them. If these changes are kept small, then the system stays close to equilibrium.