Cross-layer design for downlink multi-hop cloud radio access networks with network coding

Liang Liu, Wei Yu

Abstract

There are two fundamentally different fronthaul techniques in the downlink communication of cloud radio access network (C-RAN): the data-sharing strategy and the compression-based strategy. Under the former strategy, each user's message is multicast from the central processor (CP) to all the serving remote radio heads (RRHs) over the fronthaul network, which then cooperatively serve the users through joint beamforming; while under the latter strategy, the user messages are first beamformed then quantized at the CP, and the compressed signal is unicast to the corresponding RRH, which then decompresses its received signal for wireless transmission. Previous works show that in general the compression-based strategy outperforms the data-sharing strategy. This paper, on the other hand, points out that in a C-RAN model where the RRHs are connected to the CP via multi-hop routers, data-sharing can be superior to compression if the network coding technique is adopted for multicasting user messages to the cooperating RRHs, and the RRH's beamforming vectors, the user-RRH association, and the network coding design over the fronthaul network are jointly optimized based on the techniques of sparse optimization and successive convex approximation. This is in comparison to the compression-based strategy, where information is unicast over the fronthaul network by simple routing, and the RRH's compression noise covariance and beamforming vectors, as well as the routing strategy over the fronthaul network are jointly optimized based on the successive convex approximation technique. The observed gain in overall network throughput is due to that information multicast is more efficient than information unicast over the multi-hop fronthaul of a C-RAN.

The C-RAN Fronthaul Challenge

In a Cloud Radio Access Network (C-RAN), a Central Processor (CP) manages multiple Remote Radio Heads (RRHs) to serve mobile users. The "fronthaul"—the network connecting the CP to the RRHs—is a major bottleneck due to its finite capacity. Traditionally, two strategies handle this:

  1. Data-Sharing: The CP sends raw user messages to a subset of RRHs, which then cooperatively beamform the signal. This is essentially a multicast problem.
  2. Compression-Based: The CP calculates beamformed signals, quantizes them, and sends these compressed versions to each RRH. This is a unicast problem.

Previous research, often assuming a simple one-hop (star) topology, concluded that compression is superior. However, real-world networks are frequently multi-hop, involving various routers. This paper investigates whether the efficiency of multicasting via network coding in a multi-hop environment can tip the scales back in favor of data-sharing.

Multi-hop C-RAN architecture with CP, routers, RRHs, and users.

Methodology: Cross-Layer Optimization

The authors propose a cross-layer design that jointly optimizes the physical layer (beamforming) and the network layer (routing and network coding).

For Data-Sharing

The study leverages Network Coding. Unlike simple routing where routers just forward or replicate packets, network coding allows routers to mathematically combine data streams, achieving the theoretical multicast capacity of the network. The authors formulate a weighted sum-rate maximization problem that is non-convex and involves discrete user-RRH association decisions.

For Compression

A standard unicast routing model is used, but the beamforming vectors and the quantization noise covariance are jointly optimized.

Optimization Techniques

To solve these complex problems, Successive Convex Approximation (SCA) is employed. This technique iteratively approximates non-convex constraints with convex ones, allowing for locally optimal solutions using standard solvers like CVX. Additionally, continuous functions approximate the discrete "on/off" nature of RRH assignments (sparse optimization).

Key Results and Performance

The proposed algorithms were evaluated through numerical simulations featuring 5 RRHs and 10 users randomly distributed in a 1km radius, using an 8-node multi-hop fronthaul graph.

  • Data-Sharing Superiority: In multi-hop environments, the data-sharing strategy consistently achieved higher sum-rates than compression across various fronthaul capacities.
  • Throughput Gains: The network coding-based data-sharing scheme provided up to a 30% gain in throughput compared to schemes without optimized user-RRH association.
  • Topology Sensitivity: When restricted to a single-hop (star) topology, the compression-based strategy outperformed data-sharing by about 25%, confirming why previous studies reached different conclusions.
  • Efficiency: The SCA algorithms demonstrated fast convergence, typically reaching a locally optimal solution in fewer than 10 iterations.

Discussion: Why Multi-Hop Changes the Equation

The core insight is that in a multi-hop network, the cost of sending the same information to multiple points (multicast) is significantly lower than sending unique compressed streams to those same points (unicast). Network coding acts as a critical enabler, making the multicast routing problem polynomial-time computable and more efficient than traditional Steiner tree packing.

Furthermore, the study highlights that the superiority of a strategy is not absolute but depends on the network structure. Joint optimization is crucial because the user-RRH association acts as the "glue" coupling the radio links and backhaul links.

Conclusion and Future Directions

While this research demonstrates the potential of network coding in C-RAN, it assumes perfect Channel State Information (CSI), which is challenging in high-mobility scenarios. Future work could explore multivariate compression (jointly compressing signals for all RRHs), inter-session coding (coding across different users' messages), and robust design under imperfect CSI.

Summary updated

Abstract

There are two fundamentally different fronthaul techniques in the downlink communication of cloud radio access network (C-RAN): the data-sharing strategy and the compression-based strategy. Under the former strategy, each user's message is multicast from the central processor (CP) to all the serving remote radio heads (RRHs) over the fronthaul network, which then cooperatively serve the users through joint beamforming; while under the latter strategy, the user messages are first beamformed then quantized at the CP, and the compressed signal is unicast to the corresponding RRH, which then decompresses its received signal for wireless transmission. Previous works show that in general the compression-based strategy outperforms the data-sharing strategy. This paper, on the other hand, point s out that in a C-RAN model where the RRHs are connected to the CP via multi-hop routers, data-sharing can be superior to compression if the network coding technique is adopted for multicasting user messages to the cooperating RRHs, and the RRH's beamforming vectors, the user-RRH association, and the network coding design over the fronthaul network are jointly optimized based on the techniques of sparse optimization an d successive convex approximation. This is in comparison to the compression-based strategy, where information is unicast over the fronthaul network by simple routing, and the RRH's compression noise covariance and beamforming vectors, as well as the routing strategy over the fronthaul network are jointly optimized based on the successive convex approximation technique. The observed gain in overall network throughput is due to that information multicast is more efficient than information unicast over the multi-hop fronthaul of a C-RAN.
BibTeX
@article{liu2017cross,
  author = {Liang Liu and Wei Yu},
  title = {{Cross-layer design for downlink multi-hop cloud radio access networks with network coding}},
  journal = {IEEE Trans. Signal Process.},
  year = {2017},
  url = {https://openreview.net/forum?id=vB_-9ddjgr},
}
Figure