Link Loss Inference in Wireless Sensor Networks with Randomized Network Coding

Abstract

Due to fading and interference, data transmission via wireless links may sometimes be prone to error. For some applications in wireless sensor networks, it is of interest to monitor the link status and infer the packet loss rate. It has been shown that randomized network coding can improve the reliability of wireless sensor networks with lossy links. With network coding, the loss rate of a chosen path in a wireless sensor network is the maximum link loss rate among all the links in that path. This behavior changes the link identification problem and imposes challenges on the link loss inference. In this paper, we study the passive loss tomography problem in coded packet wireless sensor networks. We show that by inspecting the content of the coded packets at the sink (i.e., destination), one can estimate the path loss rates not only from the source nodes but also from various intermediate nodes to the sink. By utilizing such information at the sink, we determine the set of links whose loss rates can be identified. We propose a passive loss inference with random linear network coding (PLI-RLC) algorithm to estimate the link loss rates. Results show that in coded packet wireless sensor networks, our proposed algorithm can identify the status of a higher number of links compared to a Bayesian inference algorithm.