A Lightweight Energy and Spectrum Aware Decentralized Clustering Protocol for ad-hoc CR-IIoT Networks Using Trend-Based Predictive Forecasting

In ad-hoc Cognitive Radio (CR) equipped Industrial Internet of Things (CR-IIoT) networks, resource-constrained devices and stochastic spectrum availability pose significant challenge in cluster-head (CH) selection. Existing approaches rely on instantaneous observations which limits their adaptability under dynamic primary user (PU) conditions. This paper presents a lightweight decentralized clustering protocol, LESDCP, for ad-hoc CR-IIoT networks, addressing challenges due to stochastic spectrum availability and limited energy resources. Historical estimators of residual energy and idle-channel availability are incorporated using exponential smoothing, combined with Markov based forecasts and risk-aware discounting. A scalar predictive score is employed for local CH selection, eliminating centralized coordination and reducing signaling overhead. Analytical results confirm estimator stability and fairness, while simulation under dynamic PU scenarios demonstrate up to$\mathbf{6 4. 5 \%}$improvement in network lifetime and 66.7% reduction in signaling overhead compared to CogLEACH, with near-optimal bandwidth utilization is maintained.