研究目的
To propose a BBU pool resource allocation scheme in C-RAN based on recursive neural network for improving energy efficiency, reducing blocking rate, and increasing total throughput compared to traditional networks.
研究成果
The proposed LSTM-based bin-packing scheme significantly improves energy efficiency, reduces blocking rate, and increases total throughput in 5G fronthaul networks compared to traditional methods, with up to 8% improvement in throughput and zero blocking rate at certain capacities.
研究不足
The study is based on simulations and does not involve real-world deployment or hardware testing. The traffic data is limited to a specific set, and the approach may have computational complexity issues in practical scenarios.
1：Experimental Design and Method Selection:
The study uses a simulation-based approach to evaluate a proposed resource allocation scheme in a C-RAN architecture. It employs a Long Short-Term Memory (LSTM) network for traffic load prediction and a bin-packing algorithm for resource reallocation.
2：Sample Selection and Data Sources:
Traffic load data from RRHs over 21 days, averaged every 30 minutes, is used for training and testing the LSTM network.
3：List of Experimental Equipment and Materials:
No specific equipment or materials are mentioned; the simulation is computational.
4：Experimental Procedures and Operational Workflow:
The LSTM network is trained on 1008 data samples (21 days) with stochastic gradient descent, batch size of 30, and 2000 epochs. Predictions are made 30 minutes ahead, and bin-packing is applied for BBU reallocation.
5：Data Analysis Methods:
Performance metrics include number of active BBUs, power consumption, traffic throughput, and blocking rate, compared between the proposed and traditional networks.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
