Abstract
Interference management is a fundamental issue in device-to-device (D2D) communications whenever the transmitter-and-receiver pairs are located in close proximity and frequencies are fully reused, so active links may severely interfere with each other. This paper devises an optimization strategy named FPLinQ to coordinate the link scheduling decisions among the interfering links, along with power control and beamforming. The key enabler is a novel optimization method called matrix fractional programming (FP) that generalizes previous scalar and vector forms of FP in allowing multiple data streams per link. From a theoretical perspective, this paper provides a deeper understanding of FP by showing a connection to the minorization-maximization (MM) algorithm. From an application perspective, this paper shows that as compared to the existing methods for coordinating scheduling in the D2D network, such as FlashLinQ, ITLinQ, and ITLinQ+, the proposed FPLinQ approach is more general in allowing multiple antennas at both the transmitters and the receivers, and further in allowing arbitrary and multiple possible associations between the devices via matching. Numerical results show that FPLinQ significantly outperforms the previous state-of-the-art in a typical D2D communication environment.
Introduction
In Device-to-Device (D2D) networks, direct communication between nearby devices bypasses base stations to increase spectral efficiency. However, this proximity creates significant interference challenges. Managing these networks requires solving three interconnected, NP-hard problems: link scheduling (ON/OFF decisions), power control, and beamforming for MIMO systems. Existing heuristics like FlashLinQ or ITLinQ often rely on manually tuned "magic numbers" and are typically restricted to single-antenna (SISO) scenarios. This paper introduces FPLinQ, a mathematically robust framework that jointly optimizes these parameters without manual tuning.
Methodology: Matrix Fractional Programming and MM
The core of FPLinQ is the generalization of Fractional Programming (FP) to the matrix domain. While traditional FP handles scalar or vector ratios, Matrix FP enables the management of multiple data streams per link, a requirement for modern MIMO systems. The authors introduce a Matrix Quadratic Transform and a Matrix Lagrangian Dual Transform to decouple signal and interference terms within the Shannon capacity formula.
A significant theoretical contribution is the link between FP and the Minorization-Maximization (MM) framework. The paper proves that FP transforms are essentially a method for constructing surrogate functions. As illustrated in the figure above, the algorithm optimizes a simpler surrogate function that lower-bounds the complex objective, ensuring formal convergence. This approach prevents "premature turn-off," a common issue in coordinate descent where links are deactivated early and never reconsidered; FPLinQ allows links to "wake up" as the interference environment improves.
Flexible Association and Scheduling
Beyond physical layer optimization, FPLinQ addresses network topology. Most D2D models assume fixed transmitter-receiver pairs. FPLinQ introduces "Flexible Association," allowing transmitters to dynamically choose the best receiver from a set of candidates through bipartite matching.
By transforming the discrete scheduling decision into a weighted bipartite matching problem, the algorithm can find optimal configurations in polynomial time using standard methods like the Hungarian algorithm. This flexibility, combined with the lack of environment-specific thresholds, allows FPLinQ to adapt automatically to any network geometry.
Key Results and Performance
Evaluated using ITU-1411 outdoor channel models, FPLinQ demonstrates substantial gains over benchmarks:
- Throughput: In high-density environments (900 links), FPLinQ achieved nearly double the sum-rate of FlashLinQ.
- MIMO Scaling: The Matrix FP approach showed massive gains as antenna counts increased, significantly outperforming simpler vector-based FP methods by effectively managing multiple data streams.
- Fairness: When optimizing for proportional fairness (log-utility), FPLinQ provided better data rates for the bottom 50% of users compared to standard Block Coordinate Descent.
- Efficiency: The algorithm is highly convergent, with the majority of sum-rate gains realized within the first 10 iterations.
Conclusion
FPLinQ provides a unified, tuning-free optimization framework for MIMO D2D networks. By bridging Matrix FP and MM theory, it offers both theoretical convergence and superior practical performance. Future work remains to address the computational complexity for massive MIMO arrays and to develop fully decentralized versions for ad-hoc deployment.
Summary updated