A genetic approach on cross-layer optimization for cognitive radio wireless mesh network under SINR model

Due to the limited spectrum resources and the differences of link loads, how to obtain maximum network throughput through cross-layer design under signal-to-interference-and-noise ratio (SINR) model is recognized as a fundamental but hard problem. For this reason, the throughput maximization problem jointly with power control, channel allocation and routing under SINR model is researched. First, by formulating the optimization model and digging up its special structure, we show that the throughput maximization problem can be decomposed into two sub-problems: a channel allocation and power control sub-problem at the link-physical layer, and a throughput optimization sub-problem at the network layer. As to the link-physical layer sub-problem, since the joint optimization on channel allocation and power control is NP hard, we apply genetic algorithm for searching the optimal solution. As to the network layer sub-problem, we use linear programming technique for throughput optimization. To reflect the interplay property among these three layers, the fitness of each individual in the genetic algorithm is evaluated by solving the network layer sub-problem. Therefore, an effective cross-layer optimization framework based on genetic algorithm is obtained, which can find optimized power control, channel allocation and route selection in polynomial time. In order to enhance the convergence process during evolution, the integer based representation scheme and corresponding genetic operators are well designed with appropriate constraint control mechanisms. Extensive simulation results demonstrate that the proposed scheme obtains higher network throughput compared to the pervious works with comparable computational complexity.