Codebook-based uplink interference management for millimeter-wave cellular-connected uncrewed autonomous vehicle networks

Uplink transmission of Uncrewed Autonomous Vehicles (UAVs) in cellular-connected UAV networks imposes severe interference on co-channel ground users, degrading the performance of their uplink communication. To address this challenge, we propose a novel millimeter-wave (mmWave) beam tracking approach, based on users’ positions, at the Base Station (BS) side, significantly reducing mutual interference between aerial and ground users. In our proposed approach, we first design an enhanced version of the Discrete Fourier Transform (DFT) codebook, generating beams with reduced side lobes, making it suitable for multi-user Multi-Input Multi-Output (MIMO) systems and causing less interference. Subsequently, we employ a learning-based approach, Double Deep Q-Network (DDQN), at the BS side to identify the best beam directions for users, based on their location information, while planning an interference-aware trajectory for the UAV. Simulation results demonstrate that our proposed codebook outperforms well-known codebook designs, such as the IEEE 802.15.3c, beam-steering, and conventional DFT in terms of beamforming gain as the number of BS antennas increases, and signal-to-interference noise ratio (SINR) value for ground users. The performance of the beam-tracking framework using the proposed codebook in a dynamic environment, with the objective of interference mitigation, is compared with DFT codebook and Deep Q-Network (DQN). Numerical results show better performance of the proposed learning model than the DFT codebook in minimizing inter-user interference and maximizing the sum data rate of users in the shared spectrum. Besides, it shows better performance compared to DQN in selecting optimal actions, resulting in higher accumulated rewards (throughput).