Any unneeded bit transmitted from a wireless node harms the network in two ways: it wastes the power and bandwidth of the node itself, and it creates interference for other nodes. Since TCP/IP packets are on average short, existing packet headers of 40-60 bytes can create huge inefficiencies, motivating an ongoing effort in packet header compression. Unfortunately compression and error resilience are at odds: compressing the headers creates dependencies, so the compressed headers are fragile and susceptible to error propagation. This has been the central problem in packet header compression for over a decade.
This research applies powerful tools from the statistical theory of communication to the problem of packet header compression/communication. These tools have had a truly amazing track record in various problems in wireless, wireline, and optical communications, audio, image and video compression, as well as mass storage devices. This effort concentrates on the design of systems involving block codes, convolutional codes, and turbo codes and applying them to packet header compression and communication, with a direct impact on the design of better and more efficient wireless networks of the future.
