Electronic Thesis and Dissertation Repository

Thesis Format



Doctor of Philosophy


Electrical and Computer Engineering


Dr Xianbin Wang


Emergence of new applications, industrial automation and the explosive boost of smart concepts have led to an environment with rapidly increasing device densification and service diversification. This revolutionary upward trend has led the upcoming 6th-Generation (6G) and beyond communication systems to be globally available communication, computing and intelligent systems seamlessly connecting devices, services and infrastructure facilities. In this kind of environment, scarcity of radio resources would be upshot to an unimaginably high level compelling them to be very efficiently utilized. In this case, timely action is taken to deviate from approximate site-specific 2-Dimensional (2D) network concepts in radio resource utilization and network planning replacing them with more accurate 3-Dimensional (3D) network concepts while utilizing spatially distributed location-specific radio characteristics. Empowering this initiative, initially a framework is developed to accurately estimate the location-specific path loss parameters under dynamic environmental conditions in a 3D small cell (SC) heterogeneous networks (HetNets) facilitating efficient radio resource management schemes using crowdsensing data collection principle together with Linear Algebra (LA) and machine learning (ML) techniques. According to the results, the gradient descent technique is with the highest path loss parameter estimation accuracy which is over 98%. At a latter stage, receive signal power is calculated at a slightly extended 3D communication distances from the cluster boundaries based on already estimated propagation parameters with an accuracy of over 74% for certain distances. Coordination in both device-network and network-network interactions is also a critical factor in efficient radio resource utilization while meeting Quality of Service (QoS) requirements in heavily congested future 3D SCs HetNets. Then, overall communication performance enhancement through better utilization of spatially distributed opportunistic radio resources in a 3D SC is addressed with the device and network coordination, ML and Slotted-ALOHA principles together with scheduling, power control and access prioritization schemes. Within this solution, several communication related factors like 3D spatial positions and QoS requirements of the devices in two co-located networks operated in licensed band (LB) and unlicensed band (UB) are considered. To overcome the challenge of maintaining QoS under ongoing network densification and with limited radio resources cellular network traffic is offloaded to UB. Approximately, 70% better overall coordination efficiency is achieved at initial network access with the device network coordinated weighting factor based prioritization scheme powered with the Q-learning (QL) principle over conventional schemes. Subsequently, coverage information of nearby dense NR-Unlicensed (NR-U) base stations (BSs) is investigated for better allocation and utilization of common location-specific spatially distributed radio resources in UB. Firstly, the problem of determining the receive signal power at a given location due to a transmission done by a neighbor NR-U BS is addressed with a solution based on a deep regression neural network algorithm enabling to predict receive signal or interference power of a neighbor BS at a given location of a 3D cell. Subsequently, the problem of efficient radio resource management is considered while dynamically utilizing UB spectrum for NR-U transmissions through an algorithm based on the double Q-learning (DQL) principle and device collaboration. Over 200% faster algorithm convergence is achieved by the DQL based method over conventional solutions with estimated path loss parameters.

Summary for Lay Audience

Emergence of new applications and industrial automation together with the sudden increase of smart cities, smart premises (e.g., houses and buildings) and smart vehicles have led to a rapid increase in not only the number of devices connected to wireless networks but also the services provided through them. This trend has convinced the designers of the upcoming cellular wireless communication systems to make them globally available communication, computing and intelligent systems providing uninterrupted services continuously to both individuals and machines. In this kind of environment, resources used for wireless links between the devices and the serving stations are identified as extremely scarce assets that are to be very efficiently utilized. Moreover, the characteristics of these resources are specific to the locations in the 3D space as well. Based on this reason, one of the best solutions is to deviate from approximate 2-Dimensional (2D) network design principles while replacing them with 3-Dimensional (3D) network design concepts. These 3D network design principles are expected to best use the resources available at a given location in 3D space. In this case, initially a framework is developed to estimate or discover the resources at a given location under dynamic environmental conditions and to better utilize them subsequently. Mathematical modelling and self-learning techniques are used for that with the information provided by several groups of devices in a small 3D network coverage area. Coordination of both device-network and network-network interactions is also identified as one of the critical factors in the efficient utilization of these resources in meeting the Quality of Service (QoS) requirements of the devices in heavily congested future 3D networks. In this case, overall communication performance enhancement is achieved through efficient utilization of resources together with supporting techniques developed based on device and network coordination schemes. For this scenario, much attention is paid to the opportunistically available scarce resources. At the same time, factors like the locations of the devices in the 3D space, their QoS needs and priority requirements of the device congestion situations are also considered. In certain scenarios, devices served by two wireless networks are also considered where those networks are operated through different transmitters at the same location. To better facilitate these operations, prediction of receive signal power at a given location due to a transmission done by a nearby serving station is also being investigated with the objective of efficient allocation and utilization of common resources. For that, state of the art machine learning techniques are used while dynamically utilizing the limited resources.