Fully Automated Multi-Resolution Channels and Multithreaded Spectrum Allocation Protocol for IoT Based Sensor Nets

Abstract

Internet-of-Things (IoT)-based sensor networks have gained unprecedented popularity in recent years and they become crucial for supporting high data rate real-time applications. For efficient data transmission within IoT networks, it is necessary that each IoT node learns and adapts itself to recent time/spectral characteristics of channels to maximize the throughput and perform channel swapping wherever required. Many researchers have proposed channel allocation and channel quality measurement protocols within multichannel sensor networks. However, to the best of our knowledge, there is no literature available that proposes an automated and adaptive protocol that can learn and adapt according to changing channel characteristics in IoT network for achieving maximum data transmission and throughput. Therefore, this paper proposes a fully automated self-learning and adaptive protocol which can automatically transmit multiuser data by efficiently utilizing channel time/spectral characteristics. The proposed protocol is unique as it learns and adapts itself to the increasing network density based upon the network metrics. It also allows each node within IoT network to automatically detect the neighboring channel attributes so that they can swap channels to achieve maximum data transfer. This is accomplished by continuously extracting distinct features from the network topology. After extracting these features, the proposed protocol efficiently selects the best channel for an incoming node, provides the best channel utilization based upon its time/spectral attributes, and detects and allocates the unused spectrum of neighboring channels through multistage Gaussian radial basis function and multilayer perceptron-based nonlinear support vector machines classification model. Simulation results demonstrate the supremacy of the proposed protocol in terms of throughput, successful reporting probability, average blocking probability, fairness, and classification accuracy.