KRDAC: Design of a Low Complexity Routing Model for IoT Kernels Using Destination-Aware Clustering

Abstract

Internet of Things (IoT) based devices are low-power and high-efficiency motes which consist of sensors, actuators, storage units, and communication interfaces. In small-scale IoT networks, nodes directly communicate with cloud, which reduces the need of node-to-node routing. But for moderate-to-large scale networks, nodes are required to collect data from other IoT nodes in order to effectively process network information & control other devices. A wide variety of routing protocols are available for this purpose, but most of them work at application level, which incorporates inter-layer redundancies, thereby limiting their performance. Moreover, these protocols require data collection from multiple nodes, which increases their computational complexity. In order to remove these drawbacks, a kernel-level routing (KR) protocol is proposed in this text, which uses destination-aware clustering (DAC) for adhoc node grouping. The proposed protocol uses a combination of fan-shaped clustering, with distributed acyclic graphs (DAGs) in order to segregate nodes into hop-level clusters. These clusters are formed in levels, and each level is modelled such that nodes in outer level, can communicate with adjacent nodes in inner level with a single hop. The routing protocol selects nodes depending upon their instantaneous values, thereby reducing computational complexity during route selection. The proposed model was tested on different network & node configurations, and performance metrics in terms of end-to-end delay, computational complexity, energy efficiency, throughput, and packet-delivery-ratio (PDR) were evaluated. These parameters were compared with various state-of-the-art IoT routing protocols, and it was observed that the proposed KRDAC model has 8% lower energy requirements, 15% lower computational complexity, and 9% lower end-to-end delay. The proposed model was observed to be at-par with existing approaches in terms of throughput and PDR performance, thereby suggesting its large-scale deployment capabilities. This text also recommends various future research tasks which can be taken up by researchers to further improve performance of the proposed model.