Simon Krogmann,Pascal Lenzner,Alexander Skopalik,Marc Uetz,Marnix C. Vos
Abstract:We consider competitive facility location as a two-stage multi-agent system with two types of clients. For a given host graph with weighted clients on the vertices, first facility agents strategically select vertices for opening their facilities. Then, the clients strategically select which of the opened facilities in their neighborhood to patronize. Facilities want to attract as much client weight as possible, clients want to minimize congestion on the chosen facility.
All recently studied versions of this model assume that clients can split their weight strategically. We consider clients with unsplittable weights but allow mixed strategies. So clients may randomize over which facility to patronize. Besides modeling a natural client behavior, this subtle change yields drastic changes, e.g., for a given facility placement, qualitatively different client equilibria are possible.
As our main result, we show that pure subgame perfect equilibria always exist if all client weights are identical. For this, we use a novel potential function argument, employing a hierarchical classification of the clients and sophisticated rounding in each step. In contrast, for non-identical clients, we show that deciding the existence of even approximately stable states is computationally intractable. On the positive side, we give a tight bound of $2$ on the price of anarchy which implies high social welfare of equilibria, if they exist.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is about the equilibrium in the two - stage facility location game with atomic customers. Specifically, it explores how facility agents and customer agents make strategic choices to achieve their respective optimal goals on a given host graph. Different from previous research, this paper considers customers with atomic weights and allows customers to adopt mixed strategies (i.e., randomly select service facilities), which may lead to multiple different Nash equilibria in the customer sub - game. Therefore, the main problems of the research are:
1. **Existence of Pure Subgame Perfect Equilibria (SPE)**: For the scenario where all customer weights are the same, the author proves that pure subgame perfect equilibria always exist.
2. **Computational Complexity Problem**: For customers with different weights, it is computationally intractable (NP - complete) to determine whether an approximate stable state exists.
3. **Social Welfare Analysis**: Even if there are equilibria, what is the social welfare of these equilibria? The author gives a tight bound of 2 for the Price of Anarchy (PoA), which means that if there are equilibria, their social welfare is relatively high.
### Main Contributions
- **Existence Proof**: By introducing a new potential function argument method, the author proves that when all customer weights are the same, pure subgame perfect equilibria always exist.
- **Computational Complexity**: For customers with different weights, the author shows that determining the existence of an approximate stable state is an NP - complete problem.
- **Social Welfare Bound**: Gives a tight bound of 2 for the Price of Anarchy, indicating that even in the worst - case scenario, the social welfare of the equilibrium is relatively good.
### Formula Summary
The formulas involved in the paper include:
- The expected waiting time of customer \(v\), \(L_v(s, \sigma)=w(v)+\sum_{f \in N_s(v)} \sigma(s)_{v,f} \cdot \ell^{-v}_f(s, \sigma)\)
- The expected load of facility \(f\), \(\ell_f(s, \sigma)=\sum_{v \in V} \sigma(s)_{v,f} \cdot w(v)\)
- The definition of the Price of Anarchy (PoA), \(\text{PoA}=\frac{\max_{H,U,k} w(\text{OPT}(H,U,k))}{w(\text{worstSPE}(H,U,k))}\)
These formulas help quantify the impact of customer behavior, facility load, and overall social welfare.