Resource Allocation in High Dynamic Multi-UAV Networks for Coverage Continuity: A Reinforcement Learning Approach

In next-generation mobile communication system, unmanned aerial vehicles mounted aerial base stations (ABSs) are capable of providing on-demand coverage. However, resource allocation for ABSs to provide communication services with the low variance of user rate between adjacent time slots is challenging, e.g. ensuring continuous coverage, since the high dynamic of ABSs and time-varying air-to-ground channel would result in channel state information (CSI) mismatch between resource allocation decision and implementation. In consequence, the variance of user rate between adjacent time slots is large indicating that the coverage of ABSs is discontinuous in spatial temporal dimensions. To ensure the coverage continuity by minimizing the rate variance of successive time slots, we design a resource allocation method based on deep reinforcement learning (RARL). In order to adapt to the dynamic changes in network coverage requirements, the neural network structure in RARL is designed to be capable of adaptive fine-tuning. Meanwhile, to alleviate the influence of CSI mismatch between the method decision and implementation, the temporal channel correlation is taken into account in the design of reward function in RARL. Experimental results show that the rate variance of RARL can be reduced by 56.5% and spectral efficiency of RARL can be increased by 30.9% compared with benchmark algorithms, which ensures the coverage continuity.