“…While [6], [28] leveraged additional information exchange to incorporate fairness considerations into local droop-style controllers, fairness can also be imposed in centralized and distributed optimization or analytical models using equality constraints [4], [7], [29], [30] (so the same value is enforced for all end-users) or through min-max [5], [31], logarithmic [32], and quadratic [9], [33] objective function designs. On the one side, the equality constraints and the min-max objective function are hard to implement in gradient-type algorithms.…”