An Optimized Deep Learning Framework for Energy-Efficient Resource Allocation in UAV-Assisted Wireless Networks

Abstract

Unmanned aerial vehicles (UAVs) play a crucial role in resource management and offer advantages in fifth-generation (5G) and beyond networks, such as enhanced coverage, on-demand deployment, and flexible connectivity for managing data traffic spikes. However, limited power supply restricts operational time and demands efficient energy management. Real-time coordination of UAVs with ground networks and efficient data transmission further complicate resource management. Traditional methods often fail to provide the control and processing needed to optimize energy, resulting in higher operational costs, as UAVs require frequent recharging due to limited onboard energy. To address these limitations, this paper proposes a solution that combines the Double Exponential Crayfish Optimization Algorithm (DECOA) with Deep Belief Networks (DBN). DECOA merges the Crayfish Optimization Algorithm (COA) with Double Exponential Smoothing (DES) to enhance optimization for efficient resource allocation in UAV-assisted networks. DBN is applied to reduce overall network energy consumption while ensuring reliable connectivity by predicting energy consumption, enabling more accurate and optimized decision-making. The simulation results show that the proposed DECOA-based DBN model achieved an average energy efficiency of 0.80, a throughput of 0.75 Mb/s, and a computational time of 1434.5 seconds using 10 UAVs, thereby outperforming traditional methods by 26.5% in energy efficiency, 35.99% in computational time, and 35.08% in throughput. These metrics reflect a substantial reduction in energy consumption, quicker processing times, and higher data transmission rates, proving the effectiveness of the proposed DECOA-based DBN approach in UAV-assisted wireless networks.