In recent times, the concept of smart cities has been generating tremendous interest among the national governments and researchers to increase the quality of services that are offered to the citizens in a cost-efficient manner. Smart cities offer support for a wide range of applications (or services) in different domains of urban life for the betterment of the citizens. A smart city consists of a large-scale IoT network with massive amounts of IoT devices. Traditionally, the services in the smart cities are offered by utilizing the processing and storage resources of the cloud. In other words, all the data generated by the IoT devices in the smart cities are stored and processed in the cloud. This framework is not suitable for latency-critical applications like haptics, autonomous driving, Augmentative Reality (AR), etc. because the traveling latency incurred by the data to travel to the cloud is very high. To overcome this, Fog Computing (FC) has been devised where storage and processing resources are brought closer to IoT devices to minimize latency.
In this study, we propose a complete set of Topology Control (TC) techniques for constructing and managing a large-scale smart city IoT network. We approach the problem of TC in two phases: Construction Phase and Maintenance Phase. In the Construction Phase, we build a cost-effective IoT network consisting of fog gateways, and in the Maintenance Phase, we optimize the resource utilization in the system. We propose efficient algorithms in each of the phases to realize our objectives, and with extensive simulation based on real and experimented IoT data sets, we demonstrate the effectiveness of our algorithms by comparing with existing algorithms.

In the Construction Phase, compared to the existing algorithms, our Hungarian based Topology Control (HTC) algorithm performs 45% and 42% better in terms of overall system cost and the number of required gateways, respectively. In the Maintenance Phase, our Vacation based Resource Allocation (VRA) algorithm enhances the utilization of processing resources by 12x without adversely affecting the latency constraints of the applications. Similarly, our Dynamic Resource Allocation (DRA) algorithm enhances the utilization of storage resources by 4x without adversely affecting the latency constraints of the applications.