Original URL: http://www.theregister.co.uk/2013/04/23/cognitive_radio_mac_proposal/

Chinese IEEE members want MAC control for cognitive radio

Spectrum licences? Who needs 'em when radios can find their own airspace

By Richard Chirgwin

Posted in Networks, 23rd April 2013 03:18 GMT

For those in the mood for deep wireless communication geekery, a group of Chinese researchers has released a proposed protocol for MAC layer behaviour of cognitive radio systems.

The idea is to increase spectrum efficiency by allowing wireless systems to be opportunistic in their use of spectrum: that is, to use whatever appears to be “unusued” spectrum in a given time and place.

This is vastly different from today's use of spectrum. In the world of licensed spectrum, an owner (like a mobile phone operator) is granted exclusivity in spectrum and location: only one operator can use a particular slice of 900 MHz spectrum in a given location.

That operator then resolves its customers' use of that spectrum – the mobile phone users – by provisioning its network to support the expected traffic (or failing to do so, and putting up with the customer fallout).

Even so, the researchers note, no base station is active all the time, which has given rise to the proposal that “secondary users” be allowed to borrow unused spectrum during quiet periods.

For cognitive radio to work, however, low-layer protocols are needed, both to avoid the secondary users from interfering with the primary user, and to avoid two “secondary users” (SUs) shouting over each other on the same piece of spectrum – and that's why a MAC (media access control) protocol is being proposed.

In their paper, published under the IEEE imprint and available at Arxiv here, researchers Yi Liu, Shengli Xie and their collaborators propose a MAC protocol defining channel reservation, sensing, and transmission behaviours of cognitive radio.

The basis of the proposal is that all SUs have a control channel available to them, and that they compete for access to the control channel on the famous Ethernet model of “check, back off, wait a random amount of time, and check again”.

Once the control channel is reserved, the SUs would use it to communicate to each other about available spectrum for their data channel (the sensing phase), and once chosen, would use that spectrum for transmission.

The trick is to get all of this done without incurring an excessive amount of overhead. The researchers propose a variety of algorithms for this, according to different use-case examples. Their claim is that the proposed GC-MAC can at best case keep 98 percent of the available bandwidth. ®