Performance management in cognitive radio ad hoc networks using congestion based metric

In this paper, we examine performance management for cognitive radio (CR) ad hoc networks, and address the problem of data link congestion for secondary users. Our approach is to propose a metric based on the probability of contended link access and the probability of packet error to identify potential congestion on those links. We consider this metric in the context of a dynamic spectrum environment and account for effect of primary user activity on secondary user channel access. Specifically, since an ad hoc network is typically not controlled by a central base station, we propose using a clustered architecture that provides hierarchical network management utilizing a designated central manager and cluster heads that function as local managers. The cluster heads collect information from multiple layers of the protocol stack to capture the characteristics of CR ad hoc networks such as the location dependent spectrum availability. Then we define the average congestion level of the network and derive it analytically as a function of primary users´ activities and CR users´ strategy. The objective of the management action is to utilize a network level view of the congestion situation in the network by directing the CR nodes to select the highest quality links available and avoid congestion. Numerical results demonstrate the effectiveness of the proposed scheme.