Abstract
We compare the performance of standard data compression techniques in the presence of communication failures. Their performance is inferior to sending data without compression when the packet loss rate of a link is above 10%. We have developed fault-tolerant compression algorithms for sensor networks that are robust against packet loss and achieve low delays in data decoding, thus being particularly suitable for time-critical applications. We show the advantage of our technique by providing results from our extensive experimental evaluation using real sensor datasets. © 2008 Springer-Verlag Berlin Heidelberg