High-Quality and Energy-Efficient Sensory Data Collection for IoT Systems

With the advancement of sensor network technology, its application scope continues to expand. Large-scale sensor networks comprise numerous nodes capable of collecting homogeneous data from multiple sources and multiple modes. However, due to constraints on node bandwidth and energy, transmitting all data to a server would result in significant resource wastage. Furthermore, environmental noise and node failures make it challenging to ensure data reliability. Consequently, the quest for acquiring high-quality information from sensor networks while adhering to resource constraints has become an urgent issue. This paper focus on two aspects of data quality: reliability and sharing. Reliability is quantified by the deviation of data from ground truth, with smaller deviations indicating higher reliability. Sharing refers to the strong data correlation among neighboring nodes. Therefore, this paper constructs an optimization model that, under constraints related to energy and sharing, selects the most reliable data sources to transmit, maximizing the reliability of homogeneous multi-source, multi-modal data. Through experiments, genetic algorithms in sensor networks achieved a maximum improvement of 18.7% compared to the baseline in terms of data bias and a maximum improvement of 22.8% in terms of data reliability, offering an effective means for critical information acquisition in sensor networks.