Throughput-aware path planning for UAVs in D2D 5G networks

Abstract

Unmanned Ariel Vehicles (UAVs) face increasing challenges in obtaining sensory data and transferring them to the user even before the completion of their flight for time-critical processing. Traditionally bounded by only area coverage and battery capacity, UAVs now need to meet network QoS requirement when streaming data. The emergence of 5G Device-to-Device (D2D) Networks enables high speed network communication for UAVs to transfer data via D2D links during a flight. The planning of UAV flight paths is now subject to both battery capacity and network quality of service (QoS) constraints. In this paper, we focus on the path planning for UAVs, which stream data to a data receiver machine, under the constraints of full area coverage and network throughput. We present a mathematical model to formulate the issue as a combinatorial optimization problem that attempts to minimize the flight cost of multiple UAVs covering the entire area. We show that the problem is NP-hard, therefore propose a heuristic method to derive the number of UAVs and determine their flight paths. The solution is proved to be bound by $K\cdot OPT$. We conduct simulations to evaluate how the size of the area and the maximal flight distance of a UAV affect the number of UAVs needed, and how the D2D channel parameters affect the link throughputs.

Introduction

Unmanned Aerial Vehicles (UAVs) have been utilized in many critical applications including disaster relief, tracking objects, surveillance, farming, etc. In all these missions, planning the flight paths of UAVs needs to guarantee the coverage of a target region within a specific time limit, primarily due to the battery constraints [1]. In addition, the sensory data collected by a UAV are often vast in volume and highly time-sensitive. For example, UAVs used for agricultural crop survey record high resolution images [2], which are analyzed to decide on pesticide application to an infected area at specific times. The conventional “record-offload-process” approach faces a number of serious challenges such as the massive amount of sensory data and elevated risk of data loss (due to malfunction or crash of a UAV). More importantly, the time window for data processing is too short to afford offloading/processing delays because promptly treating the infected crops to prevent disease from spreading is time-critical. As a result, wireless communication capabilities are required on a UAV to transfer data continuously to a data receiver during a flight instead of after a flight.

The emergence of 5G networks brings unprecedented opportunities for high-throughput low-latency communication capabilities [3]. However, due to the shorter range of 5G radio signals especially in mmWave range, 5G base stations (aka macro or small cells) are deployed in a dense manner, attempting to support ubiquitous connectivity. As an ally to 5G network, device-to-device (D2D) communication allows devices to communicate directly without relying on cellular network infrastructure. For example, mobile User Equipment (UEs) can utilize direct links via 5G spectrum to transfer live video streams, reducing the burden on base stations [4]. In such a way, the sensing tasks by UAVs can benefit from 5G and D2D to achieve the goal of streaming data in real time.

However, the data transfer from a UAV in a wireless network environment is complex and dynamic due to the continuous movement, location change, spectrum utilization and interference. The network quality of service (QoS) is largely affected by the flight path of UAVs and how the UAVs interact with base stations and other UAVs. It is challenging to ensure the sensor data are transferred in a streaming fashion without loss to the data receiver during a flight mission. In such a context, the planning of UAV flight path remains an important task in ensuring the success of the missions, yet faces new constraints on network QoS due to data streaming requirements.

In this paper, we are motivated to study a UAV path planning problem in the context of D2D capable 5G network, where UAVs need to maintain end-to-end throughput for streaming data. The planning task is now bounded by three orthogonal constraints of network throughput requirements, battery capacity (i.e. flight time) and area coverage. Different from channel allocation problems, our focus is on how to plan flight paths to accommodate the stochastic channel behavior and achieve area coverage. Due to its continuous change in location, a UAV may not have sufficient end-to-end performance when it is distant to a base station. Therefore we aim to derive solutions that can ensure UAVs always form sufficient D2D links for streaming data.

In this paper, we first capture essential elements in this issue which is similar to a classical traveling salesman problem but with major distinctions. We define mathematically such a new path planning problem by modeling the coverage of the area and waypoints as visiting the vertices of a weighted graph. As the UAVs move, the graph turns out to be temporal and the weight (i.e. cost) of edges reflects the connectivity with D2D links. The original problem is thus transformed to a combinatorial optimization problem that aims to minimize the cost of traversal of the graph. Due to the large number of variables in the model, and the NP-hard nature of the problem, we opt to design a heuristic solution and prove its optimality bound. We conduct a simulation of UAVs flying collaboratively according to the heuristic and D2D channel model in order to evaluate how the channel parameters, area size and number of UAVs affect the D2D link throughput.

In summary, we make the following contributions in this paper:

• We present a model of the path planning problem with new network throughput constraints in a D2D 5G network. We then show that the problem is NP-hard;

• We propose a heuristic solution where UAVs collaboratively fly along a grid like pattern, and prove its approximation bound with respect to the optimal solution;

• We conduct simulations to evaluate the impact of D2D channel parameters, area size and number of UAVs on the throughput of D2D links.

The remainder of the paper is organized as follows. Section 2 discusses prior work and motivates this study. Section 3 introduces the application scenarios. We formulate the problem mathematically in Section 4. We then derive a heuristic solution in Section 6 and prove its optimality bound. Section 7 reports the evaluation results. Finally, the paper is concluded in Section 8.