An Effect of Temperature on Packet-Level Index Modulation for IoT

Real-world data collected by the Internet of Things (IoT) is essential for artificial intelligence (AI) learning. Low Power Wide Area (LPWA) systems are expected to grow with the development of IoT, because wireless systems for the IoT need to have long communication distances and be energy efficient. Packet-Level Index Modulation (PLIM) is technique that allows LPWA systems to increase the number of transmitted bits by assigning symbols to the indexes of time slots and frequency channels. However, oscillators in inexpensive IoT sensor nodes have low precision, which cause significant clock drift between the transmitter and receiver, leading to PLIM index detection errors using time-slot indexes. Existing studies have proposed methods for estimating and compensating clock drift and modeled the clock drift with a normal distribution. However, the mean of the clock drift also varies, the oscillator precisions vary with temperature. Therefore, a temperature-aware clock drift compensation algorithm with statistical-observed data has been proposed. This paper proposes a temperature-aware clock drift estimation method based on outlier detection. Machine learning could be used to detect outliers. In this paper, the errors between the measured value and the value obtained by the summary proposed method are evaluated.