Transmission Time Minimization-Based UAV Deployment and Resource Allocation With Random User Position Information

Exploiting unmanned aerial vehicles (UAVs) in terrestrial cellular networks has received considerable attention for their advanced transmission capability, flexible deployment and cost effectiveness, etc. In certain communication scenarios where the links from ground users (GUs) to base stations (BSs) or satellites may not be accessible, UAVs can be deployed as aerial relays (ARs) to forward data packets for the GUs. In this paper, we investigate the AR deployment and resource allocation problem in a UAV-assisted satellite communication system. Stressing the importance of system transmission time, we formulate the joint AR deployment, power allocation and user association problem as a constrained system transmission time minimization problem. Since the formulated optimization problem is non-convex and non-linear with the AR deployment and user association variables being coupled, it is challenging to solve. To tackle this problem, the original optimization problem is decomposed into subproblems, namely AR deployment and power allocation subproblem, and user clustering and association subproblem. Given the initial user association strategy and the number of ARs, the AR deployment and power allocation subproblem is formulated and solved by using multi-agent deep Q network algorithm. Then, given the AR deployment and power allocation strategy, we formulate the user clustering and association subproblem and propose an improvedK-means-based user clustering and association algorithm. The two subproblems are tackled in an iterative and embedded manner. Simulation results demonstrate the effectiveness of the proposed algorithms.