Optimal Resource Allocation for AGIN 6G: A Learning-Based Three-Sided Matching Approach

As ubiquitous interconnection becomes a reality for human beings, addressing the challenge of seamless coverage in the near future 6G network, particularly for remote area connections, has become increasingly urgent. HAPs and UAVs can efficiently extend the network coverage which is unfeasible by ground base stations. However, due to the high dynamic nature of air ground communications, both the available network resource and channel state information (CSI) are time varying which is typically called the dilemma of information uncertainty, often making the resource allocation intractable. To overcome the challenges, first we propose an air ground integrated 6G heterogeneous network (AGIN 6G) leveraging the wide coverage of HAPs and the hot spot communication enhancement of UAVs. Second we formulate the challenge as a three-sided matching problem with the size and cyclic preference (TMSC) among HAPs, UAVs and users. To find a feasible solution, we transform the problem into a R-TMSC issue by applying some reasonable constraints, especially when the CSI is assumed known, we can develop the HOR$<^>{2}$ A-CSI algorithm. Third, to cope with the information uncertainty, we propose the CVA-UCB matching solution augmented with machine learning. Finally, we conduct extensive experiments comparing with benchmark algorithms. Our works can verify that the proposed algorithm is superior to others in terms of the data transmission rate, system revenue and throughput.