[Paper Review] Radio Resource Allocation for Device-to-Device Underlay Communication Using Hypergraph Theory

Hongliang Zhang, Lingyang Song+1 more|arXiv (Cornell University)|Apr 12, 2016

Advanced MIMO Systems OptimizationEngineering16 references17 citations

TL;DR

This paper proposes a hypergraph-based radio resource allocation scheme for device-to-device (D2D) underlay communications to manage cumulative interference from multiple D2D pairs and cellular UEs. By modeling interference as hyperedges in a hypergraph and using a greedy hypergraph coloring algorithm, the method achieves a 33% higher cell capacity than conventional graph-based approaches, especially under high D2D pair density.

ABSTRACT

Device-to-Device (D2D) communication has been recognized as a promising technique to offload the traffic for the evolved Node B (eNB). However, the D2D transmission as an underlay causes severe interference to both the cellular and other D2D links, which imposes a great technical challenge to radio resource allocation. Conventional graph based resource allocation methods typically consider the interference between two user equipments (UEs), but they cannot model the interference from multiple UEs to completely characterize the interference. In this paper, we study channel allocation using hypergraph theory to coordinate the interference between D2D pairs and cellular UEs, where an arbitrary number of D2D pairs are allowed to share the uplink channels with the cellular UEs. Hypergraph coloring is used to model the cumulative interference from multiple D2D pairs, and thus, eliminate the mutual interference. Simulation results show that the system capacity is significantly improved using the proposed hypergraph method in comparison to the conventional graph based one.

Motivation & Objective

• To address the challenge of cumulative interference in D2D underlay communications where multiple D2D pairs share spectrum with cellular UEs.
• To overcome the limitations of traditional graph-based models that cannot capture interference from multiple sources simultaneously.
• To maximize system capacity by formulating the resource allocation problem as a hypergraph coloring problem.
• To design a low-complexity greedy algorithm with polynomial time complexity O((M+N)^3) for practical deployment.

Proposed method

• Model the D2D and cellular UEs as vertices in a hypergraph, where hyperedges represent sets of interfering transmitters affecting a common receiver.
• Define hyperedges based on cumulative interference thresholds, where a hyperedge includes all transmitters whose combined interference exceeds a predefined level.
• Formulate the channel allocation problem as a hypergraph coloring problem, ensuring no two vertices in the same hyperedge are assigned the same channel.
• Propose a greedy hypergraph coloring algorithm that assigns channels in a way that minimizes interference while maximizing the number of simultaneous transmissions.
• Use a selection threshold η to control hyperedge formation, balancing interference mitigation and system capacity.
• Integrate the hypergraph model with a channel assignment strategy that allows multiple D2D pairs and cellular UEs to share uplink channels under interference constraints.

Experimental results

Research questions

• RQ1 Can hypergraph theory effectively model cumulative interference from multiple D2D pairs and cellular UEs in underlay D2D networks?
• RQ2 How does hypergraph-based resource allocation compare to traditional graph-based methods in terms of system capacity and interference management?
• RQ3 What is the impact of varying the number of D2D pairs and available channels on the performance of the hypergraph-based method?
• RQ4 How does the choice of hyperedge size (Q) and interference threshold (η) affect system capacity and computational complexity?
• RQ5 To what extent can the proposed greedy algorithm achieve near-optimal performance with low computational overhead?

Key findings

• The hypergraph-based method achieves a 33% higher cell capacity than the graph-based method when N=50 D2D pairs, M=20 cellular UEs, and K=30 channels.
• With K=20, the D2D pair throughput is 3.0 bit/s/Hz higher using the hypergraph method compared to the graph-based method.
• The cellular UE throughput is 0.8 bit/s/Hz higher with the graph-based method at K=20, but the hypergraph method supports more UEs overall due to better interference coordination.
• The hypergraph method reduces outage probability for D2D pairs by 40% compared to the graph-based method when K=20, indicating better user scheduling.
• Increasing the hyperedge size Q from 1 to 3 improves cell capacity by about 3 bit/s/Hz at K=20, but the complexity increase outweighs the gain, justifying Q=2 as optimal.
• The system capacity saturates for D2D pairs beyond a certain density, indicating a trade-off between spectral reuse and interference.