Core Concepts
The authors study the one-dimensional Facility Assignment Game with Fair Cost Sharing (FAG-FCS) from a game-theoretical perspective, exploring equilibrium computation and strategyproof mechanism design.
Abstract
The paper investigates the one-dimensional FAG-FCS from two game-theoretical settings:
Equilibrium Computation:
In this setting, agents can directly select any facility, and the authors focus on computing a pure Nash equilibrium (PNE).
They devise a dynamic programming algorithm that computes a PNE in polynomial time, and show that the PNE attained approximates the optimal social cost within a factor of ln n.
Strategyproof Mechanism Design:
In this setting, agents report their positions to a mechanism, which then assigns them to facilities.
The authors provide a complete characterization of strategyproof and anonymous mechanisms for the case of m = n = 2, and establish a strong lower bound: no such mechanism can guarantee a bounded approximation ratio.
Inspired by the characterization, the authors design a class of non-trivial strategyproof, unanimous, and anonymous mechanisms for any n and m.
The authors demonstrate that in the one-dimensional FAG-FCS, the computation of exact PNEs is possible in polynomial time, in contrast to the general metric case. However, they also reveal a fundamental tension between strategyproofness and social cost optimization, even in the simplest congestion games.