ALGORITHMS FOR RESOURCE PLANNING AND LOAD FORECASTING IN 5G/6G MOBILE NETWORKS
DOI: 10.31673/2412-4338.2025.038718
Abstract
This article presents a comprehensive study of resource scheduling algorithms in nextgeneration mobile networks, taking into account the specific requirements of heterogeneous traffic types. A detailed analysis of classical methods, including Round Robin, Maximum C/I, Proportional Fair and their QoS-oriented modifications, is provided. It is demonstrated that traditional algorithms offer clear advantages in terms of implementation simplicity and fairness of resource allocation, but fail to effectively meet the strict requirements of ultra-reliable low-latency communication (URLLC) and enhanced mobile broadband (eMBB) services. The paper also explores modern approaches based on machine learning and deep reinforcement learning, which demonstrate high adaptability and flexibility in 5G/6G scenarios, but require significant computational resources and large training datasets.
Special attention is devoted to traffic load control and prediction methods, which have evolved from reactive mechanisms such as Admission Control and Load Balancing towards intelligent systems utilizing Big Data analytics, cognitive radio, and software-defined networking (SDN/NFV). The study shows that ARIMA models provide basic capabilities for short-term traffic forecasting, while LSTM-based neural networks allow for capturing both temporal and spatial dependencies, thus improving prediction accuracy in dense urban 5G/6G scenarios.
An experimental study was conducted using a simulation of a mobile network with 60 users and 10 resource blocks, enabling the evaluation of various scheduling algorithms under dynamic channel and traffic conditions. The results were visualized through throughput and delay graphs, confirming the expected behavior: Maximum C/I achieves the highest throughput but neglects fairness, Proportional Fair strikes a balance between efficiency and stability, M-LWDF effectively minimizes delay, while a DRL-like scheduler provides balanced performance across multiple metrics. The findings suggest that hybrid approaches, combining classical algorithms with intelligent methods, are the most promising solution to ensure high quality of service and efficient resource utilization in current and future mobile networks.
Keywords: mobile networks, resource scheduling, algorithms, QoS, load prediction, 5G, 6G, machine learning, reinforcement learning, URLLC, eMBB, mMTC
