Efficient Management of Resource-Constrained UAV Swarms in Critical Applications

Abstract

Unmanned aerial vehicles (UAVs) have attracted significant attention due to their advantages over traditional techniques in critical applications, such as surveillance, search and rescue operations, provision of wireless communication services to ground users, and last-mile package delivery. Their ability to access remote or hazardous areas while minimizing human risk has paved the way for innovative solutions in diverse fields. In critical applications, the efficient management of resource-constrained UAV swarms is imperative to ensure the high efficiency of utilizing UAVs in their missions. However, adopting UAVs in critical applications such as surveillance and post-disaster scenarios presents a myriad of challenges that demand careful consideration. In surveillance-based scenarios, end-to-end latency and limited available resources onboard UAVs pose critical challenges for UAVs to complete their mission successfully. Addressing the challenge of latency in communication is essential, as low-latency data transmission is crucial for surveillance-based applications. Moreover, UAVs' constrained computation and memory capabilities pose significant hurdles in critical applications that require real-time data processing and decision-making. Another critical application is utilizing UAVs to provide services in post-disaster situations. Several challenges need to be addressed to ensure the efficient use of UAVs in post-disaster applications. Among these challenges is the inherent limitation of energy resources available onboard UAVs. Striking a delicate balance between optimizing mission duration and minimizing energy consumption necessitates sophisticated energy-aware path planning algorithms and alternative energy harvesting solutions. The maximization of the throughput of data collected from distributed Internet of Things (IoT) devices on the ground further complicates matters, requiring intelligent and advanced algorithms to guide the UAVs throughout their missions. Achieving optimal coverage and connectivity in wireless communication systems demands adaptive swarm formations, adding complexity in dynamic and unpredictable environments. The maximum flight duration of UAVs, restricted by onboard power, poses a critical constraint, especially in scenarios with prolonged mission requirements. Lastly, adhering to strict time deadlines for collected data in critical applications demands real-time processing capabilities and adaptive algorithms. Addressing these multifaceted challenges is pivotal for the successful deployment of resource-constrained UAV swarms in critical applications where efficiency, reliability, and responsiveness are paramount. Overcoming the challenges of UAVs in critical applications demands efficient management of their available resources. Efficient management of UAV swarms not only ensures UAV mission success but also enhances the utilization of the limited UAV resources, reduces operational costs, and extends the overall lifespan of the UAVs. Advanced algorithms enable UAVs to navigate the covered area, optimizing energy consumption, energy harvesting, wireless communication coverage, throughput of the data collected from IoT devices, and end-to-end latency. Addressing the challenges of obtaining precise and real-time solutions for efficient management of UAVs resources encounters significant obstacles due to the rapid environmental changes and the complexity of solving nonlinear equations. State-of-the-art approaches such as optimization-based, game-theoretic-based, and heuristic-based approaches are expected to face substantial constraints in addressing the efficient resource management of UAVs problems due to their limited scalability, inadequate adaptability, high computational demands, restrictive information exchange, and reliance on solving nonlinear equations. Consequently, AI-driven solutions have emerged as a transformative solution to tackle these challenges across diverse scenarios. In particular, recent advancements in deep reinforcement learning (DRL) techniques have emerged as promising and practical techniques to navigate UAVs path planning issues within wireless networks. DRL-based methods for UAVs efficient resource management have demonstrated efficient outcomes, outperforming or achieving comparable performance to conventional approaches requiring high complexity due to solving nonlinear equations. This research study provides practical solutions for managing resource-constrained UAV swarms in two main scenarios. The first scenario focuses on utilizing UAVs for surveillance applications, optimizing end-to-end latency in distributed inference of neural networks while considering limited UAV resources. The primary goal is to ensure reliable data transmission among UAVs, employing a novel approach that exploits onboard resources to locally make classification decisions, eliminating the need for data relay to remote servers. The second approach is utilizing UAVs to provide services in post-disaster scenarios. In particular, the UAVs aim to deliver wireless power transfer (WPT) in the downlink channel and data collection from distributed IoT devices on the ground in the uplink channel. This approach aims to optimize critical parameters, including maximizing energy harvesting from IoT devices, data throughput using techniques like reflecting intelligent surfaces (RIS), and coverage of the distributed user elements (UEs) on the ground. It considers practical constraints such as the maximum UAV flight time, the time deadline for collected data, the UAV energy consumption, and the minimum throughput for successful transmission. Finally, our work aims to maximize wireless connectivity coverage to scattered IoT devices on the ground and throughput of collected data by utilizing multiple RIS technologies.

Date of Award	2024
Original language	American English
Awarding Institution	HBKU College of Science and Engineering