Abstract
Power control is a widely used technique in improving the throughput and the quality of services in wireless communications. In this paper, a power control method based on a utility maximization method is applied to Ad-hoc networks built on Wi-Fi radios. First, a fully distributed power control algorithm is proposed to handle the discrete nature of Wi-Fi data rates. This is then integrated with a greedy scheduling mechanism to ensure fairness and prevent node starvation in the network.
Introduction
In Wi-Fi ad-hoc networks, mutual interference between nodes significantly degrades performance. While Network-wide Utility Maximization (NUM) is a standard theoretical framework for addressing this via power control, it faces two major practical hurdles. First, standard optimization often leads to "bang-bang" control, where nodes causing high interference are simply shut down to maximize aggregate throughput, sacrificing individual fairness. Second, real-world Wi-Fi radios operate with discrete "staircase" data rates rather than continuous functions, making the optimization problem non-convex and computationally expensive.
Methods: The JPSA Framework
The proposed Joint Power control and Scheduling Algorithm (JPSA) addresses these challenges through a two-tiered distributed approach that combines power control with autonomous scheduling.
Distributed Power Control
Each node independently calculates its optimal transmission power using a modified Lagrangian method. Nodes account for the interference they cause to others by incorporating a "price" ($\lambda$) into their local utility function. To accommodate the discrete rate sets of commodity Wi-Fi hardware, a staircase utility function is employed alongside a gradient update mechanism designed to prevent divergence and ensure stable convergence.
Greedy Scheduling and Fairness
To resolve the "fairness trap" where power control might zero out a node's power to benefit the network average, JPSA incorporates a greedy scheduling mechanism. When a node is identified as needing to be turned off to maintain network stability, the algorithm triggers a "channel split." Nodes are autonomously grouped into subgroups that utilize Time Division Multiple Access (TDMA) to share the channel. This ensures that every node receives a transmission slot while maintaining high overall efficiency.
Experimental Validation
The algorithm's assumptions were validated using the ORBIT wireless testbed at Rutgers University. Measurements with Atheros-based Wi-Fi cards confirmed that the Signal-to-Interference-plus-Noise Ratio (SINR) model is applicable to real hardware. Specifically, the team demonstrated "power capture" effects where stronger packets can be decoded despite collisions, grounding the theoretical model in physical reality.
Results and Discussion
Simulation results across 5,000 random topologies demonstrate that JPSA achieves linear complexity $C(N)$, making it suitable for real-time implementation on mobile devices. This is a significant improvement over traditional joint scheduling/power algorithms which often exhibit exponential complexity.
In a specific 5-link test scenario, pure power control maximized throughput at 84 Mbps but completely starved two links. In contrast, JPSA maintained high aggregate throughput while ensuring all five links remained active. As shown in the iteration plots, the algorithm converges quickly, stabilizing power levels and prices while maximizing the sum rate without sacrificing individual link connectivity.
Conclusion
JPSA provides a computationally efficient, distributed solution for interference management in Wi-Fi ad-hoc networks. By treating scheduling as a "release valve" for power control, it balances the competing goals of network-wide throughput and individual node fairness. While implementation currently faces hurdles due to closed-source MAC layer drivers, JPSA demonstrates the feasibility of applying advanced utility maximization to commodity wireless hardware.
Summary updated