CoMP in the Sky: UAV Placement and Movement Optimization for Multi-User Communications

TL;DR

This paper introduces a novel UAV-based wireless network architecture employing CoMP for interference mitigation, optimizing UAV placement and movement over time to enhance multi-user uplink communication throughput.

Contribution

It develops a new multi-UAV CoMP system model with mobility-aware placement optimization using random matrix theory for rate bounds, a novel approach in UAV communication networks.

Findings

UAV placement is a weighted average of user locations and previous/future UAV positions.

Derived tight bounds for user achievable rates under LoS channels.

Optimized UAV movement improves network throughput and fairness.

Abstract

Driven by the recent advancement in unmanned aerial vehicle (UAV) technology, this paper proposes a new wireless network architecture of \emph{coordinate multipoint (CoMP) in the sky} to harness both the benefits of interference mitigation via CoMP and high mobility of UAVs. Specifically, we consider uplink communications in a multi-UAV enabled multi-user system, where each UAV forwards its received signals from all ground users to a central processor (CP) for joint decoding. Moreover, we consider the case where the users may move on the ground, thus the UAVs need to adjust their locations in accordance with the user locations over time to maximize the network throughput. Utilizing random matrix theory, we first characterize in closed-form a set of approximated upper and lower bounds of the user's achievable rate in each time episode under a realistic line-of-sight (LoS) channel model with random phase, which are shown very tight both analytically and numerically. UAV placement and movement over different episodes are then optimized based on the derived bounds to maximize the minimum of user average achievable rates over all episodes for both cases of full information (of current and future episodes) and current information on the user's movement. Interestingly, it is shown that the optimized location of each UAV at any particular episode is the weighted average of the ground user locations at the current episode as well as its own location at the previous and/or next episode. Finally, simulation results are provided to validate and compare the performance of the proposed UAV placement and movement designs under different practical application scenarios.