Graduation Year


Document Type




Degree Name

Doctor of Philosophy (Ph.D.)

Degree Granting Department

Electrical Engineering

Major Professor

Yasin Yilmaz, Ph.D.

Committee Member

Nasir Ghani, Ph.D.

Committee Member

Ismail Uysal, Ph.D.

Committee Member

Ankit Shah, Ph.D.

Committee Member

Lu Lu, Ph.D.


5G Cellular Networks, Edge Computing, Intelligent Vehicular Systems, Resource Allocation, Smart Cities, Ultra-Reliable Low-Latency Communications


In view of the recent advances in Internet of Things (IoT) devices and the emerging new breed of smart city applications and intelligent vehicular systems driven by artificial intelligence, fog radio access network (F-RAN) has been recently introduced for the next generation wireless communications. The capability of F-RAN has emerged to overcome the latency limitations of cloud-RAN (C-RAN) and assure the quality-of-service (QoS) requirements of the ultra-reliable-low-latency-communication (URLLC) for IoT applications. To this end, fog nodes (FNs) are equipped with computing, signal processing and storage capabilities to extend the inherent operations and services of the cloud to the edge. However, due to their limited resources compared to the cloud, FNs should utilize their valuable resources intelligently to satisfy various QoS requirements in synergy and complementarity with the cloud. This dissertation considers the network slicing problem of sequentially allocating the limited resources at the network edge (i.e., FNs) to vehicular and smart city users with heterogeneous latency and computing demands in dynamic environments.

The network slicing problem for the edge resource allocation is formulated as a Markov decision process (MDP), for which dynamic programming and reinforcement learning (RL) solutions are proposed to adaptively learn the optimal slicing policy. The research conducted in this dissertation can be divided into two phases. Phase I considers a single FN with a tractable number of states, where tabular solution methods are used such as dynamic programming and model-free RL methods, which include Monte Carlo, SARSA, expected SARSA, and Q-Learning (QL). Phase II deals with a cluster of FNs coordinated with an EC with high-dimensional state space, where deep RL (DRL) methods such as deep Q-networks (DQN) are adopted to address the high-dimensionality using deep neural networks. A network slicing model based on a cluster of FNs coordinated with an edge controller (EC) is developed to efficiently utilize the limited resources at the network edge. Specifically, for each service request in a cluster, the EC decides which FN to execute the task (i.e. locally serve the request at the edge) or to reject the task and refer it to the cloud to conserve the edge resources for future users of potentially higher utility to the system (i.e., lower latency requirement).

The developed network slicing model and the proposed RL and DRL algorithms quickly learn the optimal policy through interaction with the IoT environment, which enables adaptive and automated network slicing for efficient edge resource allocation in dynamic traffic and load profiles of dense vehicular and smart city service requests with heterogeneous latency and computing needs. This means providing a cost-effective service customization through virtual partitioning of the RAN resources to enable low-latency IoT communications such as autonomous driving and public safety operations. The proposed approach also adapts to various design objectives including edge resource utilization, cloud offloading, and satisfying diverse QoS requirements.