الفهرس | Only 14 pages are availabe for public view |
Abstract Internet of things (IoT) technology enables interconnections among a tremendous number of things in both urban and rural areas. Mobile Edge Computing (MEC) is a network architecture that enables cloud services to be hosted on edge devices near the users. Empowering IoT with MEC can essentially improve the QoS for IoT applications. However, MEC resources increase the energy consumption and cost of the system. Moreover, the MEC system needs an orchestrator to schedule IoT tasks on MEC resources and ensure that the QoS of all tasks is satisfied. IoT in rural areas also faces a problem in delivering tasks to the MEC devices due to the limited network coverage. In this thesis, we suggest a solution to these challenges for heterogeneous applications in urban areas and non-time-sensitive applications in remote areas. In the urban setting, we propose a resource design optimization algorithm and a partially centralized two-level cooperative scheduling algorithm with dual-threshold server state control. The resource design optimization algorithm uses discrete particle swarm optimization (PSO) to distribute resources on MEC devices such that system availability is satisfactory at minimum cost. The scheduling algorithm allocates resources to tasks based on the deadline and size of incoming tasks. State control is proposed to reduce the energy consumption of the MEC system by deactivating unused computation resources. A dual threshold control policy is used to reduce state switches when traffic fluctuates, therefore stabilizing the system. The threshold values that balance energy minimization, system stability, and system availability are obtained over two steps using discrete PSO. In the rural setting, we propose a framework for designing an energy-efficient UAVassisted data collection framework for non-time-sensitive stationary IoT applications. In our framework, data collection occurs over two steps. First, IoT devices send the collected data to a nearby aggregator over low-energy channels. The position of aggregators is optimized using a triangulation-based clustering method that minimizes the number of aggregators needed to decrease the system cost and energy consumption. Then the data collected by aggregators are passed to UAVs that relay the data to the internet. The energy consumption of the second stage data collection is minimized by optimizing the location of the UAV dockstation, the communication power, and the UAV trajectories. UAV dockstation position and the communication power are optimized using gaining-sharing knowledge (GSK) metaheuristic, while UAV trajectories are optimized through solving a capacitated vehicle routing problem (CVRP). |