Learn-As-You-Fly: A Distributed Algorithm for Joint 3D Placement and User Association in Multi-UAVs Networks

Abstract

In this paper, we study the joint 3D placement of unmanned aerial vehicles (UAVs) and UAVs-users association under bandwidth limitation and quality of service constraint. In particular, in order to allow to UAVs to dynamically improve their 3D locations in a distributed fashion while maximizing the network's sum-rate, we break the underlying optimization into 3 subproblems where we separately solve the 2D UAVs positioning, the altitude optimization, and the UAVs-users association. First, given fixed 3D positions of UAVs, we propose a fully distributed matching based association that alleviates the bottlenecks of bandwidth allocation and guarantees the required quality of service. Next, to address the 2D positions of UAVs, we adopt a modified version of K-means algorithm, with a distributed implementation, where UAVs dynamically change their 2D positions in order to reach the barycenter of the cluster that is composed of the served ground users. In order to optimize the UAVs altitudes, we study a naturally defined game-theoretic version of the problem and show that under fixed UAVs 2D coordinates, a predefined association scheme, and limited-interferences, the UAVs altitudes game is a non-cooperative potential game where the players (UAVs) can maximize the limited-interference sumrate by only optimizing a local utility function. Therefore, we adopt the best response dynamics to reach a Nash equilibrium of the game which is also a local optimum of the social welfare function. Our simulation results show that, using the proposed approach, the network's sum rate of the studied scenario is improved by 200% as compared with the trivial case where the classical version of K-means is adopted and users are assigned, at each iteration, to the closest UAV.

Energy-Efficient Trajectory Optimization for UAV-Assisted IoT Networks

“…Second, a locally optimal solution with the constraints, including maximum speed and the users' energy neutrality, is achieved by alternating optimization and successive convex programming. In [32], the authors propose a distributed algorithm that allows UAVs to maximize the network's sum rate by dynamically learning the optimal three-dimension (3D) locations associated with ground users. The algorithm decomposition breaks the optimization into three subproblems addressed by a distributed matching-based association, a modified version of the K-means algorithm, and a game-theoretic algorithm with a local utility function.…”

New Trends in Stochastic Geometry for Wireless Networks: A Tutorial and Survey

“…A comprehensive survey about channel modeling for UAVassisted communications can be found in [174]. Also, given their technical constraints combined with ground UEs QoS requirements, optimal placement of UAVs is another challenging task, which may include UAVs trajectory optimization [259], altitude optimization [173], [260], flight time optimization [259], [261], and UAVs density optimization [262].…”

Learning in the Sky: Towards Efficient 3D Placement of UAVs

“…where d u,u and d min represent the 3D distance between UAV u and UAV u , and a certain minimum distance, respectively. According to (8), if the distance between UAV u and UAV u is less than a minimum distance d min , the function Γ u (t) returns value one which is considered as a cost in the utility function defined in (7). Let A(t) = (a 1 (t), .…”

“…Most research efforts have addressed this issue from a non-learning perspective [2]- [4]. However, there has been a growing attention recently devoted to the use of learning algorithms for the deployment problem of UAVs [5]- [8]. In [5] and [6] learning based approaches to find the two-dimensional (2D) trajectory of UAVs flying with fixed altitudes were proposed.…”

“…However, these works do not address the 3D deployment of multiple UAVs integrated into already existing terrestrial networks. In [8], the problem of 3D placement was dealt as a two separate optimization problems in horizontal 2D plane and altitude of UAVs, where a prior knowledge of the users locations is required. However, providing this information for UAVs in real time can be challenging.…”