Abstract
This paper explores the mobile association problem in wireless networks using the mathematical framework of Optimal Transport Theory. By modeling user demand as a continuous fluid rather than discrete entities, the authors characterize optimal cell boundaries for two distinct objectives: minimizing total network power (centralized optimization) and achieving a stable state for selfish users (decentralized equilibrium). The study reveals that traditional Voronoi-based association is often suboptimal under varying scheduling policies and non-uniform user densities.
Introduction
In modern wireless networks, the association between mobile terminals (MTs) and base stations (BSs) is a critical factor in determining overall performance. While the standard approach often defaults to connecting a user to the nearest base station—resulting in a geometric Voronoi partition—this method fails to account for network-wide power efficiency, interference, or varying user densities. This paper addresses the mobile association problem by determining the optimal service area, or "cell," for each base station under different operational regimes.
System Model and Mathematical Framework
The authors employ a fluid approximation, treating the user population as a continuous probability density function $f(x, y)$ over the network area. This allows the application of Optimal Transport Theory (specifically the Monge-Kantorovich problem) to find smooth cell boundaries. The connection between users and BSs is framed as a "transport map," where the goal is to "move" data from BSs to users in the most efficient manner.
The notation used throughout the analysis is summarized in Table 1.
Optimization Scenarios
The paper analyzes two primary scenarios:
- Global Optimization (Centralized Case): A central operator aims to minimize the total power consumption of the entire network while satisfying the data requirements of all users.
- User Optimization (Decentralized Case): Intelligent mobile terminals independently choose their BS to maximize their own throughput. This leads to a Wardrop Equilibrium, a concept from traffic flow theory where no user can improve their throughput by unilaterally switching BSs.
The analysis further considers two specific scheduling policies: Round Robin, where each user receives an equal time fraction, and Rate Fair, where each user is guaranteed a specific average transmission rate regardless of their distance from the BS.
Key Insights and Results
A significant finding is that the classic Voronoi strategy is only optimal under the Rate Fair policy in the absence of capacity constraints. Under other policies, such as Round Robin, optimal cell boundaries shift significantly to balance load and power.
In the centralized case, a "power-density tradeoff" emerges: if users are highly concentrated near one BS, the network may expand the cell of a neighboring BS to offload traffic and minimize total power, even if those users are physically further from the second BS. Conversely, in 2D non-uniform scenarios, high user density can lead to increased interference, causing a central BS's cell to shrink to maintain a viable Signal-to-Interference-plus-Noise Ratio (SINR).
The paper also identifies a "Price of Anarchy"—a performance gap between centralized and decentralized scenarios. Selfish throughput maximization by individuals often leads to network-wide power inefficiencies or unfairness compared to the global optimum.
Conclusion
By applying Optimal Transport Theory to wireless networks, this research provides a robust mathematical foundation for determining optimal cell boundaries. The results demonstrate that geometric association is frequently suboptimal when considering power consumption and user fairness. Future work will extend this framework to dynamic user distributions and more complex radio environments, such as LTE architectures.
Summary updated