UAV Positioning with Joint NOMA Power Allocation and Receiver Node Activation

This paper proposes reinforcement learning (RL)-based solutions for unmanned aerial vehicle (UAV) data offloading in B5G mmWave-enabled communications. This is particularly useful for ad-hoc transmission scenarios within environments experiencing connectivity issues with the main servicing network as in disaster-stricken areas. Double deep Q-network and multiarmed bandit-based algorithms are proposed to tackle the joint problem of UAV-positioning and Rx-node activation and power allocation for data offloading in downlink NOMA transmissions. Numerical simulations are performed to ensure the proposed RL-based algorithms can adequately provide high data transfer rates, along with random and exhaustive search solutions as benchmarks for lower and upper bounds on the achievable sum-rate levels.