Deep Learning Anomaly Detection for Cellular IoT With Applications in Smart Logistics

The dataset was generated using NB-IoT edge nodes. We created a setup where an edge node has been attached to a box-shaped container inside a transport vehicle moving through the city of Novi Sad. The devices were initially connected to the NB-IoT network, and they had the uninterrupted connectivity along their paths. We collected the positioning data from GNSS module (timestamp, latitude, longitude, altitude, speed and number of satellites in range), as well as the outputs of the IMU (acceleration and magnetic field along the 3 spatial axes). The time resolution (sampling period) of the GNSS samples was approximately 10 s. The sampling period of the IMU was approximately 15 ms. We calculated the RMS as well as the arithmetic mean for the acceleration and magnetic field samples collected within a GNSS sampling interval.