Wireless research projects


Mobile phone and TV networks are carefully planned in advance, to place masts efficiently and provide uniform coverage. But many modern wireless networks have no central planning: people just buy routers and set them up where they like, causing problems with interference, loss of connectivity, and security. These chaotic networks are increasingly common in modern communication networks. The ERC-funded CHAOSNETS project is a ground-up redesign of chaotic wireless networks, focusing on making them more scalable, secure, and reliable. We are investigating fountain coding and receiver-based rate adaptation methods to improve wireless capacity in the vagaries of the grey zone of marginal coverage, improvements to security and localisation based on the profiling of incoming signals' angles of arrival access points, and interference mitigation and management techniques for networks not under the same administrative control.

Software radio

In standard wireless hardware, digital signal processor (DSP) chips take in a wireless signal, synchronize it, demodulate it, decode it, and present the answer to the software. With a software radio, all of the signal processing is done in software. This lets us build network protocols that use radio in new ways.

For example, today's network protocols run congestion control algorithms to slow down their transmission rate when the network is congested, and they detect congestion by dropped packets. What if the software radio could tell the congestion controller how much radio interference there is, so that the congestion controller can back down smoothly, e.g. reducing its transmission rate and increasing the redundancy in its coding, so that no packets need be dropped?

Cone of Silence: Cooperative Interference Mitigation for WiFi Networks

WiFi's cheap hardware and license-free operation have led to tremendous crowding of unlicensed bands. In urban environments where WiFi is deployed densely, interference severely limits users' throughput. In Cone of Silence (CoS), we are building a practical WiFi system that greatly mitigates this problem by harnessing multiple antennas to intelligently null toward users in nearby competing networks, thus reducing the interference they experience. Unlike prior WiFi beamforming systems, a CoS access point intelligently nulls toward those nodes in competing networks that will benefit most from a reduction in interference, while prioritizing preservation of the receive throughput of its own clients. And unlike prior such systems, it does so in purely decentralized fashion, without requiring trust or coordination over a wired connection between competing WiFi base stations.