Wireless sensor networks (WSNs) are a key technology in the big data age. Deployed sensors, equipped with wireless communication capabilities, collect massive data from remote or even hostile environments. They transmit the data to a collection point known as the sink. Energy efficiency and operational life are important topics in WSN research. It is impractical and sometimes impossible to change the batteries for these sensors. Where do we strike the balance between sufficient sampling and energy conservation?

The authors start from the beginning and explain the use cases of WSNs and the different models of data collection. The paper focuses on the time-driven model where collection and transmission take place on a periodic basis. This type of WSN is also known as a periodic sensor network (PSN). The sampling rate of the sensors can be set on high to capture the intensive data changes in critical events, or on low to avoid oversampling and save energy. What if the PSN can adaptively adjust its sampling rate to the demand of events? The authors present three statistical tests (Fisher, Turkey, and Bartlett) and discuss how to use them to gauge the variance of data. Furthermore, the authors formulate a behavior function to model the adaptation. At the end, the paper numerically demonstrates the instantaneous sampling rates and energy consideration of these three tests.

Step by step, the authors introduce new concepts and debate the rationale behind design decisions. The theories and formulas are well defined and explained. Even for readers not familiar with WSNs, the presentation is clear and organized enough to guide them through the discussion. The only disappointment in my opinion is the lack of realistic experiments. Even though the paper mentions utilizing the data from a real testbed, it remains unclear how the data set fits into the adaptive model for validation.