Measurement-Calibrated Distributed Gateway Selection for Canopy-Aware Forest STIN-IoT Networks
Abstract
In this paper, reliable and energy-efficient data collection is investigated for canopy-aware forest satellite-terrestrial integrated Internet of Things (STIN-IoT) networks, where direct sensor-to-satellite transmission is constrained by long-distance propagation, vegetation-induced attenuation, limited non-terrestrial network (NTN) uplink resources, and costly battery maintenance. To address these issues, a hierarchical architecture is developed in which battery-powered sensor nodes transmit sensing data to nearby satellite-capable gateway terminals through low-power wide-area network (LPWAN) access links, while selected gateway terminals aggregate and forward the collected data through NTN uplinks. A canopy-aware two-stage reliability-energy model is established to jointly characterize terrestrial access, satellite forwarding, vegetation attenuation, data aggregation, queue evolution, and device energy consumption. To reduce the mismatch between analytical NTN link models and practical forest operation, field measurements collected from a solar-powered satellite IoT gateway are used to calibrate the gateway-to-satellite outage probability. Based on the calibrated model, a slot-level joint gateway-selection and sensor-association problem is formulated, and a measurement-calibrated distributed gateway selection with burden-aware rotation (MC-DGS-BR) algorithm is proposed. Each candidate gateway locally computes a suitability score according to its residual energy, access reachability, measurement-calibrated NTN outage risk, queue backlog, and accumulated burden. An overlap-aware election rule suppresses redundant gateway activation, while the burden-aware rotation mechanism mitigates repeated gateway usage and the energy-hole effect. The main contribution is a measurement-calibrated distributed decision framework that connects field-observed NTN reliability with long-term energy-balanced gateway operation without requiring global network-state collection. Simulation results demonstrate that MC-DGS-BR achieves lower effective outage probability and higher energy efficiency than the random-selection and without-selection benchmark schemes under different forest deployment and channel conditions.