| Wireless mesh network (WMN) is one of the most promising next-generation wire-lessnetworktechnologies, whichhasshoweditsadvantagesindeployment, coverage, andbandwidth and so on. Routing protocol is a key component of WMN. Due to the specificsof WMN, such as multi-hop transmission, mesh inter-connection, multi-gateway coexis-tence, distributed deployment and configuration, quasi-static mesh points, equipped withadvanced wireless devices, etc., existing routing protocols for other multi-hop wirelessnetworks lose their optimality in WMN. Therefore, it is critical to conduct comprehensiveresearch in designing new routing protocols for WMN. Multi-path routing, cross-layerdesign and network coding are three potential technologies toward the success of WMNrouting protocols. This dissertation offers deep dive into those technologies as follows.Probing-based online anypath routing (POAR) protocol and expected online any-path transmission time (EOATT) metric are proposed to address shortcomings of existingWMN single-rate single-channel multi-path routing protocols, such as inferior transmis-sion throughput and delay, unbalanced traffic distribution, inefficient network resourceutilization and so on. By mapping the WMN routing problem to a Stochastic Recover-able Canadian Traveler Problem (SRCTP), POAR effectively resolves local link failureproblems caused by transient link uncertainties. POAR applies a centralized candidateforwarder selection mechanism, which remarkably reduces the forwarding decision over-head. Evaluation results from a dedicated simulator based on NS-2 reveal that, comparedwithsemi-deterministicrouting,POAReffectivelyreducesend-to-enddelayandimprovespacket delivery ratio. By using RTS/CTS as the online probing tool in NS-3 simulator,POAR has both significant goodput improvements and much less packet jitter.Probing-based online multi-rate anypath routing (POMrAR) protocol and expectedonline multi-rate anypath transmission time (EOMrATT) metric are proposed to addressshortcomings of existing multi-rate WMN routing protocols, such as suboptimal rate se-lection and inferior transmission throughput. By mapping the multi-rate WMN routingproblem to a stopping problem, on the ground of optimal stopping rule from stopping the-ory, POMrAR thus can choose the best forwarding time aimed at optimizing end-to-endtransmissiondelayinmulti-rateWMN. AlsoPOMrARnaturallydegradestoPOARundertwo-state link model. Evaluation results from a dedicated simulator based on NS-2 revealthat, compared to multi-rate semi-deterministic routing, POMrAR reduces end-to-end de-lay and improves packet delivery ratio.Probing-based online multi-channel anypath routing (POMcAR) protocol and ex- pected online multi-channel anypath transmission time (EOMcATT) metric are proposedto address shortcomings of existing multi-channel WMN routing protocols, such as poortrade-off between channel diversity and network connectivity, unbalanced gateway over-head and inferior transmission throughput. Based on POAR protocol, POMcAR employsa two-phase channel setting procedure, achieves better trade-off between channel diver-sity and network connectivity, and improves transmission throughput in multi-gatewayWMN. Based on NS-3 simulation, inside network component, POMcAR has comparablethroughput with ROMA protocol, which is much higher than the common channel proto-col. Significant improvements on transmission throughput are also observed across net-work components. Furthermore, compared with single path routing, POMcAR achievesbetter throughput.Weighted partial network coding (WPNC) and its enhancements are proposed to ad-dress shortcomings of traditional network coding, such as high decoding delay and receiv-ing jitter, unawareness of packet weights. A generalized framework of applying WPNC inmulti-path routing in WMN is also discussed. WPNC is based on partial network coding(PNC). By replacing the full square coefficient matrix in network coding with a right up-per triangle matrix, WPNC not only inherits many characteristics of network coding suchas simplifying multi-path packet scheduler and improving transmission reliability throughredundancy, butalsoinheritsalladvantagesofPNC,suchaspartialdecodingabilitywhichcan effectively reduce decoding delay and jitter. Moreover, WPNC supports weight basedencoding, decoding and data replacement operators. With collect more or smart node en-hancements, WPNC can easily reach 100% decoding probability. Theoretical analysisfor WPNC has been conducted via MATLAB numeric evaluations, which further prove theadvantages of WPNC.
|
PDF Full Text Request