Intelligent Resource Optimization in Integrated Aerial Terrestrial Networks

Awais, Muhammad and Pervaiz, Haris and Ni, Qiang and Yu, Wenjuan (2025) Intelligent Resource Optimization in Integrated Aerial Terrestrial Networks. PhD thesis, Lancaster University.

Text (2025MuhammadAwaisPhD)
2025MuhammadAwaisPhD.pdf - Published Version
Available under License Unspecified.
Download (2MB)

Abstract

Developing sixth generation (6G) wireless communication necessitates low-power consumption, high-reliability, and massive-connectivity. One of the most promising solutions to address these requirements is aerial base station (ABS) based communication systems that employ both in-the-air (aerial) and on-the-ground (terrestrial) components. ABSs enhance line of sight (LoS) connections, fulfilling escalating quality of service (QoS) demands. Nevertheless, integrating aerial and terrestrial networks into future three dimensional (3D) networks introduces emerging requirements for resource allocation and new functional challenges, such as latency, reliability, energy consumption, and QoS. Motivated by the above observations, this thesis investigates the challenges of intelligent resource optimization in integrated aerial terrestrial networks. An integrated aerial and terrestrial network is initially examined to design a bisection-based low-complexity adaptation (BLCA) algorithm for optimal resource allocation. A joint optimization problem that involves sub-carrier (SC) assignment, blocklength, and power allocation (PA) subject to delay, reliability, and QoS constraints is investigated to enhance system performance in a finite blocklength (FBL) regime. The proposed solution includes sub-carrier allocation based on matching theory, optimal blocklength allocation using the bisection algorithm, and a two-step projected gradient descent power distribution by optimizing the power budget on each sub-carrier. Case studies on flexible blocklength and PA are also examined under the next generation of multiple access techniques. The second part integrates digital twin (DT) technology with mobile edge computing (MEC) to facilitate mobile offloading in an integrated aerial-terrestrial network. An advanced bisection sampling-based stochastic solution enhancement (BSSE) algorithm is specifically tailored to jointly optimize transmit power, central processing unit (CPU) frequency, and the task offloading policy to minimize the system's energy-time cost against benchmarks. The proposed solution includes a one-climb policy to narrow the search space, a closed-form solution for calculating the optimal CPU frequency and transmit power for given offloading decisions, and an inequality condition formulated to manage dependent tasks efficiently. The scalability of the proposed scheme is also analyzed. In the final part, machine learning techniques are adopted to improve the system performance in an integrated aerial-terrestrial wireless network. The proposed solution employs unsupervised learning techniques for the grouping of internet of things smart devices (ISDs), Q-learning (a type of reinforcement learning) for the intelligent ABS placement, and deep learning for power allocation. A closed-form expression is also derived for PA among multiplexed devices based on their QoS requirements. Numerical results indicate that the proposed scheme significantly outperforms existing benchmark schemes. This thesis presents valuable insights into innovative, sustainable, and energy-efficient resource optimization in integrated aerial-terrestrial future-generation networks, setting the stage for further advancements in resource allocation to enhance reliability, QoS, and latency.