Abstract
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 HAIM. The data required as input to the training phase is captured experimentally using IEEE 802.15.4 compliant Crossbow MICAz motes. We demonstrate the RAM-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.