Novel Intelligent Sector-Based Approach for Congestion and Interference-Aware Load Balancing In 5G Heterogeneous Networks

Abstract Over the years, the unpredictable nature of user activities and traffic patterns has often caused overload and poor overall quality of service in 5G networks. The aim of this paper is a novel intelligent sector-based approach for congestion and interference-aware load balancing in 5G Heterogeneous Networks (HetNet). The research gap is the lack of an improved sector-based approach with congestion and interference awareness. The methodology used is experimental and simulation. The research design began with data collection of 5G HetNet traffic information from MTN Nigeria. The 5G network considered is HetNet with three cells, which are macro, micro, and pico. The area considered is Enugu Sub-urban. The joint congestion and interference problem was mathematically formulated as a non convex mix integer non-linear programming problem. The collected data were processed, and then applied to train five Machine Learning (ML) models (Multi-Layer Perceptron (MLP), Random Forest (RF), Support Vector Machine (SVM), Extreme Gradient Boosting (XGB), and Logistic Regression (LR)). To validate the models, five Deep Learning (DL) models (Convolutional Neural Network (CNN), Long Short Term Memory (LSTM), CNN+LSTM, Gated Recurrent Unit (GRU) and Bi-directional GRU (Bi-GRU)) were also trained and there results all compared considering metrics such as Coefficient of Determination (R2), Mean Square Error (MSE), and Root Mean Square Error (RMSE)). From the results, the XGB recorded the best performance with MSE of 0.0008, RMSE of 0.0291 and R2 of 0.9966. Sector-based control was proposed using Clear Channel Assessment (CCA) and Carrier Annulling Algorithm (CAA). The intelligent traffic prediction model was applied to improve the proposed sector based for load balancing. The CAA utilized signal to noise ratio to sense interference in the channels while CCA was applied to avoid resource allocation to busy channel. Collectively these models were integrated to curate the novel intelligent sector-based congestion and interference-aware solution for 5G HetNet. In this paper, we call it the Intelligent Sector Based (ISB). The ISB was integrated on the 5G HetNet and simulated and compared against the benchmark Static Sector Based (SSB). The results reported 31.5% improvements for congestion management against the SSB, and also 55.9% noise reduction when compared to SSB. More simulations were done considering different user activities which include live streaming, phone calls, and file transfer. The results obtained showed consistent network performance success with ISB. Existing recent models in the literature like Proportional Fair (PF) and Deep Reinforcement Learning (Deep RL) were also compared with the ISB and ISS. The results obtained revealed that the model competes closely with the best which is Deep RL; however, the ISB is the best model which has the capacity to control both congestion and interference with a 5.905% improvement against Deep RL considering signal-to-noise ratio. In conclusion, this ISB is recommended for congestion and interference management in 5G HetNet.