Enhancing Bandwidth Allocation Efficiency in 5G Networks with Artificial Intelligence

The explosive growth of data traffic and heterogeneous service requirements of 5G networks—covering Enhanced Mobile Broadband (eMBB), Ultra-Reliable Low Latency Communication (URLLC), and Massive Machine Type Communication (mMTC)—present tremendous challenges to conventional methods of bandwidth allocation. A new deep reinforcement learning-based (DRL-based) bandwidth allocation system for real-time, dynamic management of 5G radio access networks is proposed in this paper. Unlike rule-based and static strategies, the proposed system dynamically updates itself according to shifting network conditions such as traffic load and channel conditions to maximize the achievable throughput, fairness, and compliance with QoS requirements. By using extensive simulations mimicking real-world 5G scenarios, the proposed DRL model outperforms current baselines like Long Short-Term Memory (LSTM), linear regression, round-robin, and greedy algorithms. It attains 90%–95% of the maximum theoretical achievable throughput and nearly twice the conventional equal allocation. It is also shown to react well under delay and reliability constraints, outperforming round-robin (hindered by excessive delay and packet loss) and proving to be more efficient than greedy approaches. In conclusion, the efficiency of DRL in optimizing the allocation of bandwidth is highlighted, and its potential to realize self-optimizing, Artificial Intelligence-assisted (AI-assisted) resource management in 5G as well as upcoming 6G networks is revealed.