Research Of Cross-layer Optimization Technologies Based On State Sensing And Error Compensation In Wireless Mesh Network

Wireless mesh network (WMN), which is known as a special broadband wirelessnetwork structure, has been widely applied due to its particular advantages, such as fastconstruction, low upfront costs, large coverage area, good scalability, etc. However, thecharacteristics of WMN itself, such as time-varying channel, unstable topology, limitedresources of nodes and link bandwidth and so on, bring in uncertainty to the netowrk, andmight even make the wire wideband network protocol stack unavailable. To address theseproblems, cross-layer design is proposed. With cross-layer design, the network stateinformation can be exchanged between different layers and different layers can even bejointly modeled and designed. And in this way, each layer can detect the change of networkstatus quickly, adaptively adjust the resource allocation in order to satisfy different QOSrequirements and eventually realize the optimal resource allocation.Until now, most of researches on cross-layer resource allocation algorithm are based onsuch assumptions that the network is static, and network state information is perfect andtime-invariant. Actually, the network is dynamic due to the time-varying channel, mobility ofnodes, feedback delay and so on. Based on it, the dynamic cross-layer resource allocationalgorithm for WMNs is further studied in the paper. With consideration of dynamiccharacteristic, using error compensation, a series of cross-layer resource allocation algorithm,which can quickly detect and automatically adapt to network changes, is proposed for WMNs.The main contributions are as follows:(1) Using loosely coupled cross-layer design, a cross-layer multipath link state routingprotocol WHMRA, which can not only capture the network congestion and interference, butalso adapt to the network changes quickly, is proposed based on careful analysis on thecharacteristics of WMN and the limitation of the existing method. It differs from thetraditional routing protocol as follows:a) Using loosely coupled cross-layer design, multiple interfaces are designed betweennetwork layer, MAC layer and physical layer and a new routing metric is obtained. Comparedwith existing ones, on one side, the new metric can capture the inter-flow/intra-flowinterference and traffic load; on the other side, it is isotonic so that low computationcomplexity algorithm can be used to compute the efficient routing for data packets.b) By extending the coverage-based Multipoint Relay (MPR) selection algorithm to alink-quality-based MPR selection algorithm, the routing protocol makes a tradeoff betweenrouting quality and routing overhead. The simulation results show that the link-quality-based MPR selection algorithm can make network layer find a good-quality path and improvesystem throughput.c) By combining the link-quality-based MPR selection mechanism and fuzzy slightedphilosophy, WHMRA can use lower overhead to detect the changes of physical layer statussuch as traffic, topology quickly, and then coordinate the routing configuration. Simulationresults show that WHMRA can adapt to the topology change, reduce the routing overhead,and improve the routing scalability.(2) Using tightly coupled cross-layer design, a SINR-based conditional average capacityis obtained with the consideration of outdated CSI and a joint optimal congestion control,channel allocation and power control algorithm for WMN with outdated CSI is proposed. Thedetails are described below:a) Due to the limitation of spectrum resource, frequency reuse in space is usuallyadopted to improve system capacity of wireless networks, so that interference occurs betweendifferent links in the effective jamming range. Interference links increase especially in amulti-hop environment. According to this, it is a relatively complex process to compute theSINR-based conditional average capacity for WMN. By firstly obtaining the probabilitydensity function of current channel gain on outdated channel gain, the closed-form expressionfor the expectation of conditional capacity under SINR model can be further derived througha multiple integration process.b) The impact of outdated CSI on cross-layer resource allocation is firstly considered indynamic multi-hop wireless networks and a framework with outdated CSI is proposed tojointly optimize congestion control in the transport layer, channel allocation in the data linklayer and power control in the physical layer. In the proposed framework, the network ismodeled as a generalized network utility maximization (NUM) problem with elastic link datarate and power constraints and consequently, the NUM problem is solved in both a centralizedand a distributed manner. Both two are convergent, and can compensate the error of resourceallocation and improve the resource utilization efficiently.(3) Using tightly coupled cross-layer design, an analytical framework is proposed for theoptimization of network performance through joint congestion control, channel allocation,rate allocation, power control, scheduling and routing with the consideration of fairness andoutdated CSI in the multi-channel WMN.a) The joint congestion control, channel allocation, rate allocation, power control,scheduling and routing problem is modeled as a complex non-linear mixed integerprogramming problem. Due to the computation complexity, the centralized algorithm is impractical especially for a mobile network. According to this, using the dual decompositiontechnique, the primal problem is decomposed into several subproblems:1) flow control intransport layer,2) next-hop routing,3) rate allocation and link scheduling in data link layer,4)power control in physical layer and5) channel allocation in MAC layer, and finally thesesubproblems are solved by the coordination of different dynamic network state parameterssuch as congestion price and so on. The algorithm keeps the convergence, and significantlyoutperforms the existing algorithms on network utility and energy efficiency.b) Considering the time-varying channel and feedback delay of CSI, the research resultsin a) are extended by channel prediction technology, the error brought by outdated CSI iscompensated and the correctness of resource allocation algorithm in a dynamic environmentis improved. Simulation results demonstrate that the prediction-based algorithm compensatesthe performance cost brought by outdated CSI and shows better performance than the existingones.