Cross-layer Optimization For Multihop MIMO Cognitive Radio Networks

Multiple-input multiple-output (MIMO) technology is an effective means tobreak Shannon capacity constraints. Without the need to expand the spectrum, MIMOtechnology uses multiple interfaces and multiple channels to simultaneously transmitdata that can exponentially increase the system capacity of wireless multi-hopnetwork, improve spectrum utilization efficiency and save spectrum resources.Meanwhile, MIMO technology has a significant impact on multi-hop wirelessnetwork throughput, real-time connectivity and topological properties. So we cancombine physical layer, MAC layer and the transport layer network to study cross-layer optimization. The cognitive radio (CR) embedded with intelligence informationprocessor which can perceive the external spectrum resources, CR can flexibly adjustthe transmit and receive parameters based on the sensed information so that seconduser (SU) can be assigned dynamic primal users (PU) spectrum resources, so CRtechnology can improve the network performance. The integration of MIMO and CRtechnologies can significantly improve the existing network congestion condition, ithas broad applications, so it is important to study in cross-layer optimizationtechnology based on MIMO and CR technologies.Based on lots of research on the current situation of this area, we proposedistributed algorithms to solve the proposed cross-layer optimization for multi-hopMIMO cognitive radio network problem. We made the following research:(1) We propose a green multi-hop MIMO cognitive radio network under flatfading model in ad hoc network, which has node transmit power constraint and linkcapacity constraint. We use dual decomposition algorithm to decompose the originalproblem into the network-transport layer and the physical-link layer sub-problems.Network-transport layer sub-problem is a convex problem; physical-link layer isnon-convex problem. For the network-transport layer sub-problem, we use thedistributed Newton algorithm to solve it. For the physical-link layer sub-problem, wepropose the linearization-based alternating algorithm to solve it. For the main problem,which is non-convex, this may lead to non-feasible solution to the original problem,so we propose an effective heuristic method to recover to a feasible solution.(2) Based on research (1), we propose a multi-hop MIMO cognitive radio network under the frequency selective fading model. We consider the interferencebetween the SUs, the problem has more practical value and becomes more complex.Similarly, we use dual decomposition methods to decompose the original probleminto the network-transport layer and the physical-link layer sub-problems. Thenetwork-transport layer sub-problem is the same with the problem in chapter two. Forthe physical-link layer non-convex sub-problem, it is equivalent to the weight andMSE minimization problem (WMMSE), we use the linear transceiver designalgorithms to solve WMMSE. For the main problem, which is non-convexity, thismay lead to non-feasible solution to the original problem, so we also propose aneffective heuristic method to recover to a feasible solution.(3) The network-transport layer sub-problem of research (1) and research (2) is aconvex optimization problem. Distributed Newton method can be used to solve theproblem. In order to enhance the scope of distributed Newton’s method, we expandthe network transport layer sub-problem to the wireless sensor network problem.Wireless sensor networks applications, its performance mainly based on routing andflow control. We design a distributed Newton method with quadratic convergenceperformance to maximize the network utility. We apply matrix splitting technique todistributedly solve Newton dual variables through single-hop information exchange.