DRL-Based Resource Allocation and Trajectory Planning for NOMA-Enabled Multi-UAV Collaborative Caching 6G Network

Aerialplatform-based network has been invoked as an appealing assistance for terrestrial cellular networks. Caching at edge UAVs is an effective emerging solution for relieving the wireless backhaul congestion and reducing content fetching latency for hotspot areas. This article investigates a non-orthogonal multiple access (NOMA)-enabled multi-UAV collaborative caching network, where the multimedia contents are delivered from UAVs to ground users. UAVs cache popular contents and collaboratively share with each other. The optimization problem of joint caching decision, 3D trajectory planning and spectrum resource allocation is formulated for minimizing the system content retrieving delay. It is an NP-hard mixed integer nonconvex optimization issue with extra difficulty that the content popularity is highly dynamic. To tackle this challenging problem, we decouple the original issue into two subproblems: (1) UAV collaborative caching decision making and (2) joint optimization of UAV trajectory planning, power allocation, and channel reusing. We develop a multi-agent proximal policy optimization (MAPPO)-based algorithm for the former to achieve the maximum weighted content hit ratio and propose a matching-DRL solution for the latter. Numerical evaluations confirm that the proposed approach is capable of rapid convergence and achieves superior performance in terms of content hit ratio, content retrieving delay, and system throughput compared to other benchmark algorithms.