Cluster channel equalization using adaptive sensing and reinforcement learning for UAV communication

Abstract

Aiming to address the need for dynamic sensing and channel equalization in UAV cluster communication environments, this article introduces an equalization algorithm based on a U-Net model and fuzzy reinforcement Q-learning (U-FRQL-EA). This algorithm is designed to enhance the channel sensing and equalization capabilities of UAV communication systems. Initially, we develop a U-Net-based signal processing algorithm that effectively reduces acoustic noise in UAV communication channels and enables real-time, accurate perception of channel states by automatically learning channel features. Subsequently, we enhance fuzzy reinforcement Q-learning by incorporating a fuzzy neural network to approximate the Q-values and integrating this approach with the allocation strategy of wireless sensing nodes. This enhancement not only improves the accuracy of Q-value approximation but also increases the algorithm’s adaptability and decision-making ability in complex channel environments. Finally, we construct the U-FRQL-EA equalization algorithm by combining the improved U-Net model with fuzzy reinforcement Q-learning. This algorithm leverages the U-Net model to sense channel states in real time and intelligently adjusts data forwarding strategies based on fuzzy values generated by the fuzzy reinforcement Q-learning. Simulation results demonstrate that the U-FRQL-EA algorithm effectively reduces the system’s bit error rate, enhances communication quality, and optimizes network resource utilization, offering a novel solution for improving the performance of uncrewed aerial vehicle communication systems.

Keywords

• Blind equalization
• U-Net
• Fuzzy reinforcement Q learning
• Adaptive perception
• Dynamic equalization