We identify how to share costs locally in weighted congestion games with polynomial cost functions in order to minimize the worst-case price of anarchy (PoA). First, we prove that among all cost-sharing methods that guarantee the existence of pure Nash equilibria, the Shapley value minimizes the worst-case PoA. Second, if the guaranteed existence condition is dropped, then the proportional cost-sharing method minimizes the worst-case PoA over all cost-sharing methods. As a byproduct of our results, we obtain the first PoA analysis of the simple marginal contribution cost-sharing rule, and prove that marginal cost taxes are ineffective for improving equilibria in (atomic) congestion games.
展开▼