Quality of Service-driven Channel Selection for Cognitive Radio Networks (VENI-CR)

Themes: Sensor network communication

Disaster relief networks suffer from spectrum clogging. This is visible in the Netherlands, where during large disasters emergency teams cannot communicate reliably (viz. the Turkish Airlines crash near Schiphol in 2009). To solve this problem a Cognitive radio (CR) is needed. In case of insufficient capacity CR-enabled devices detect and operate on the unutilized licensed radio channels.

The fundamental component of CR disaster relief network is the channel selection mechanism. Based on metrics like channel time-domain statistics (e.g. mean availability time), CR devices select the radio resource that satisfies Quality of Service (QoS) requirements such as message delivery rate. While physical layer metrics used for QoSbased channel selection are well studied, use of time-domain metrics in channel selection is an open scientific problem in CR. As the collection of data on channel utilization must be done sporadically/irregularly to minimize delay, incautious estimation might in-turn hinder the operation of CR network.

In disaster relief networks, data flows have strict QoS requirements, also expressed in terms of the maximum allowed delay or minimum throughput. I postulate that channel selection has to be adapted to these QoS goals and depend on confidence levels on estimated traffic parameters. The key idea is to exploit the information on the estimated distribution of licensed user traffic in channel selection process. Features like variance or skewness of the estimated traffic distribution would help in quantifying data flow QoS disruption probability by instantaneous licensed user appearance. Primary innovations include (i) investigating fundamental limits of licensed traffic estimation accuracy, given time and physical layer constraints; and (ii) devising core methods of channel selection based on distributed parameter estimation considering QoS requirements.