Modeling, estimating and predicting the packet-level Bit Error Rate process in IEEE 802.15.4 LR-WPANs using Hidden Markov Models

This paper describes a stochastic wireless channel model that captures the behavior of the packet-level bit error rate (BER) and the link quality indication (LQI) processes. The model is based on a discrete-time hidden markov model (HMM) whose hidden states correspond to different BERs, and whose observable states correspond to different LQI values. We use the Baum-Welch algorithm to train the HMM. The data required as input to the training phase is captured experimentally using IEEE 802.15.4 compliant Crossbow MICAz motes. We demonstrate the HMM-based channel model´s (HCM) utility and versatility by three applications. In the first we use it to synthesize traces whose properties closely resemble those of the training data. This application simultaneously demonstrates the HCM´s correctness as well. In the second application we demonstrate the HCM´s ability to estimate a received packet´s BER based on the LQI with which it was received. In the third application we demonstrate the HCM´s ability to predict the BER to which future packets will be subjected. For evaluation purpose we restrict our prediction to the next packet.