Decentralized Throughput Maximization in Cognitive Radio Wireless Mesh Networks

• Published in: IEEE transactions on mobile computing Vol. 13; no. 9; pp. 1967 - 1980
• Main Authors: El-Sherif, Amr A., Mohamed, Amr
• Format: Magazine Article
• Language: English
• Published: Los Alamitos IEEE 01.09.2014; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithm design and analysis; Algorithms; Allocations; Approximation; Availability; Cognitive Mesh Networks; Cognitive radio; Delay; Distributed Resource Optimization; Economic models; Interference; Links; Mathematical models; Mathematical Software; Mathematics of Computing; Maximization; Mesh networks; Networks; Nonlinear approximation; Numerical Analysis; Probability and Statistics; Queuing theory; Resource management; Scheduling; Sensors; Spectrum allocation; Throughput; Wireless networks; mesh networks; distributed algorithms; Resource allocation; cognitive radios
• ISSN: 1536-1233; 1558-0660
• DOI: 10.1109/TMC.2013.82

Scheduling and spectrum allocation are tasks affecting the performance of cognitive radio wireless networks, where heterogeneity in channel availability limits the performance and poses a great challenge on protocol design. In this paper, we present a distributed algorithm for scheduling and spectrum allocation with the objective of maximizing the network's throughout subject to a delay constraint. During each time slot, the scheduling and spectrum allocation problems involve selecting a subset of links to be activated, and based on spectrum sensing outcomes, allocate the available resources to these links. This problem is addressed as an aggregate utility maximization problem. Since the throughput of any data flow is limited by the throughput of the weakest link along its end-to-end path, the utility of each flow is chosen as a function of this weakest link's throughput. The throughput and delay performance of the network are characterized using a queueing theoretic analysis, and throughput is maximized via the application of Lagrangian duality theory. The dual decomposition framework decouples the problem into a set of subproblems that can be solved locally, hence, it allows us to develop a scalable distributed algorithm. Numerical results demonstrate the fast convergence rates of the proposed algorithm, as well as significant performance gains compared to conventional design methods.