Leveraging uncertainty for transmission period control in IoT applications

In numerous Internet of Things (IoT) applications, vast amounts of data are continuously collected and transmitted by IoT sensors to enable precise, real-time environmental monitoring. However, frequent data transmission significantly increases energy consumption, posing a critical challenge for IoT systems. To address this issue, we propose the uncertainty-based optimal transmission period control (U-OTPC) model, which adaptively adjusts the transmission period of each IoT sensor to minimize energy consumption while preserving data accuracy. The optimal transmission period is derived by formulating a min–max optimization problem that balances energy consumption and data quality, where data quality is quantified using a combination of reconstruction error and uncertainty. To estimate predictive uncertainty, we leverage the Monte Carlo (MC) dropout technique, a factor often overlooked in transmission period control models. To efficiently solve this problem, the U-OTPC integrates theoretical propositions with a k -ary search algorithm. The effectiveness of the proposed model is rigorously validated through extensive evaluations on three distinct open datasets collected from real-time monitoring. Experimental results show that the U-OTPC model surpasses other benchmarks in the period control score (PCS) metric, effectively balancing data collection accuracy and energy consumption. • We newly introduce an uncertainty-based optimal transmission period control (U-OTPC) model. • The proposed model incorporates MC dropout to quantify the uncertainty of the prediction model. • We formulate a min–max optimization problem considering energy consumption and data quality. • We present a sophisticated U-OTPC algorithm that employs k -ary search to achieve an optimal solution. • Extensive evaluations show our U-OTPC model defeats the conventional models in the PCS metric.