Theorem 3, PDF p. 5 — on m = 2 identical links every Nash equilibrium has social cost ≤ (3/2)·opt
ProvedWorstCaseEq.Identical.theorem3_price_of_anarchy_two_linksLet agents with positive traffics each send their traffic over one of two identical parallel links, choosing the link at random with probabilities , independently of each other. Let be a Nash equilibrium: no agent can lower its expected cost, the expected traffic on the link it uses, by changing its probabilities unilaterally. Let
be the social cost, the expected maximum traffic over the two links, and let be the least maximum link traffic over all assignments of the agents to the links. Then
This is Theorem 3 of the paper: the price of anarchy (coordination ratio) for any number of agents and two identical links is at most . Together with the instance of Theorem 1 (two unit agents, each uniform over the two links), which attains , it determines the price of anarchy on two identical links exactly.
Formalization Note "The price of anarchy is at most " (a bound on over equilibria) is stated as the inequality for every equilibrium, which avoids a quotient with . Positive traffics are part of the paper's model. The equilibrium is the full mixed Nash condition of the published agt_games vocabulary, payoffs being minus costs. Links are Fin 2; is allowed and then both sides are .
import Mathlib import Definitions.Def_agt_games import Definitions.Def_WorstCaseEq_Identical_Model
namespace WorstCaseEq.Identical
/-- Theorem 3, PDF p. 5: for any number of agents and `m = 2` identical links, every Nash equilibrium has
social cost (expected maximum link traffic) at most `3/2` times the optimum. -/
theorem theorem3_price_of_anarchy_two_links {n : ℕ} (w : Fin n → ℝ) (hw : ∀ i, 0 < w i)
(p : Fin n → Fin 2 → ℝ) (hp : IsNash w p) :
socialCost w p ≤ (3 / 2 : ℝ) * opt 2 w := by sorry
end WorstCaseEq.Identical
Confirmed by the mission captain (proposal self-audit).
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.