UAV-Assisted wireless communications

Unmanned aerial vehicles (UAVs) have been attracting a lot of attention in recent years for its potential in numerous applications. Due to the flexibility and mobility of UAVs, UAV-mounted base stations are effective and cost-efficient to provide wireless connectivity and to improve the performance of terrestrial wireless network when fixed infrastructure is not available. In particular, UAVs can timely adjust their locations according to the movement of ground users (GUs). Against this background, we firstly propose a method to apply mobile edge caching on UAVs in wireless communication systems. By investigating the user request preference with the aid of latent Dirichlet allocation (LDA), the caching strategy can be optimized. In the proposed system, we consider the design of intelligent caching strategies when a number of UAVs are deployed to serve the GUs, where each UAV has a limited storage capacity for caching useful user contents. We use LDA to extract the user request preferences in order to intelligently caching data in the UAVs, while the k-means clustering is utilized to classify GUs and to assist in deploying the UAVs. Additionally, we consider three user-UAV association criteria, namely the user received signal to noise ratio (SNR), user preferences and the delay. Our simulation results show that, when compared to random caching, the average caching efficiency could be significantly improved from 50% to 70%, while the latency of our proposed system can also be greatly reduced. Then, we propose an optimized UAV-user association technique that can mitigate the challenge of the existence of “outliers” caused by limited UAV resources. More specifically, in order to solve the sudden-shift of UAVs position, which is a challenge for traditional deployment algorithm such as the k-means clustering, we propose a reinforcement learning based method to eliminate the challenge of sudden-shift requirement as well as to provide an improved performance, when we have limited UAV resources. Our simulation results show that the proposed UAV-user association can provide a solution for the “outliers” problem. Meanwhile, the average downlink rate of the system, which employs the proposed deep Q-learning (DQL) based UAV deployment, is close iv to the ideal system that employs a k-means clustering with infinite UAV speed. Additionally, we show that the DQL provides an improved performance, when the number of UAV is limited for a given coverage area. On the other hand, given the limited availability of on-board energy, we design an energy efficient communication scheme. Explicitly, we propose a method that combines the concept of index modulation (IM) with the UAV communications systems, which we refer to as IM-UAV, to attain an improved energy efficiency (EE). Furthermore, based on the proposed IM-UAV communication system, a gradient descent based UAV deployment scheme is designed to maximize the sum rate of the GUs in the target area. Additionally, the maximum likelihood detection for the IM-UAV requires a high computational complexity for detection at the GUs, while providing the best possible performance. Hence, we propose a low-complexity detection scheme that can separately detect the index symbols and data symbols to reduce the computational complexity at the receiver side. The simulation results demonstrate that the proposed deployment method is capable of attaining the appropriate positions to deploy the UAVs, while the EE is also improved by combining IM with UAV communication system. Finally, we consider a content-aware scenario that employs UAVs as aerial base stations to transmit data to GUs via air to ground communication links. Furthermore, inspired by the concept of rate splitting (RS), we design two subcarrier-sharing (SS) data transmission schemes to improve the average data rate of the GUs. Additionally, based on the proposed transmission schemes, we design the UAV deployment schemes, namely the fixed-point deployment scheme and traverse-search deployment scheme. The simulation results demonstrate that our proposed data transmission techniques significantly increase the average data rate, while the deployment schemes are capable of optimizing the data rate in the respective application scenarios