DeepHopping-LTE: A Deep Reinforcement Learning Framework for Adaptive Frequency Hopping Spread Spectrum (FHSS) in Dense 4G Cellular Networks

The rapid growth of mobile devices and dense 4G deployments has intensified interference and congestion, challenging network reliability, spectrum efficiency, and quality of service. This study proposes DeepHopping-LTE, a deep reinforcement learning-based adaptive frequency hopping framework designed to dynamically optimize frequency selection in dense cellular networks. By leveraging a deep Q-network (DQN), the framework balances exploration and exploitation using an epsilon-greedy policy, converging to optimal hopping sequences to minimize interference while maximizing throughput. Simulations and field data from two urban locations demonstrated that DeepHopping-LTE significantly outperforms traditional static FHSS and heuristic frequency hopping methods. The framework achieved a stable signal-to-interference-plus-noise ratio (SINR) averaging 15–17 dB, compared to 5 dB for static methods, more than doubling spectrum efficiency to 5.5 bits/s/Hz under high load conditions. Packet delivery ratio also improved dramatically, ranging from 94% to 98%, while traditional FHSS averaged 75%, reflecting superior reliability and quality of service. Throughput improvement ratios reached 1.6 under peak user density, and interference reduction ranged from 40% to 70%, confirming effective mitigation of environmental noise. Additionally, the adaptability index reached 0.95, demonstrating near-instantaneous response to dynamic network changes, whereas heuristic and static methods lagged at 0.65 and 0.15, respectively. The results validate that reinforcement learning can transform frequency hopping from a static approach into an intelligent, adaptive solution capable of optimizing network performance in real-time. By continuously observing network states and selecting optimal channels, DeepHopping-LTE not only mitigates interference but also enhances spectrum utilization, reliability, and user experience. These findings provide a practical foundation for policy implementation in dense urban networks, supporting guidelines for adaptive frequency management, interference monitoring, and reinforcement learning deployment in 4G and future 5G environments. In conclusion, DeepHopping-LTE addresses the critical challenges of dense cellular networks, offering a scalable, intelligent, and resilient framework that surpasses traditional methods, enabling efficient and reliable communication in high-density mobile ecosystems.