Robust Over-The-Air Aggregation for uplink OFDM system under burst sparse interference

Over the air computation (AirComp) is proposed as a vital technology in physical layer for big data applications in internet-of-things (IoT) systems. However, interference may severely affect the aggregation process. In this work, we propose an analog product code and present a low complexity Bayesian decoder for analog error correction in IoT-AirComp in the presence of burst sparse impulsive interference. We model the burst interference using a Markov model and cancel it via streaming variational message passing based Bayesian inference algorithm. The Bayesian approach allows us to learn global parameters directly via measurements in a streaming fashion, thus facilitating online machine learning. We provide extensive simulations to verify the model and show improved performance against small as well as burst sparse impulsive interference.