Autoencoder-Based Characterisation of Passive IEEE 802.11 Link Level Measurements

Wireless networks are indispensable in today's industrial manufacturing and automation. Due to harsh signal propagation conditions as well as co-existing wireless networks, transmission failures resulting in severe application malfunctions are often difficult to diagnose. Remote wireless monitoring systems are extremely useful tools for troubleshooting such failures. However, the completeness of data captured by a remote wireless monitor is highly dependent on the temporal, e.g., short-term interference, and spatial characteristics of its environment. It is necessary to first ensure that the data was completely captured at the remote monitor in order to maintain the integrity of the failure analysis, i.e., to avoid false positives. In this paper, we propose an autoencoder-based framework to evaluate the quality of wireless data captured at a remote wireless monitor. The algorithm is trained using data generated under controlled laboratory conditions and validated on testbed as well as realworld measurement data.