Research On Domain-oriented Fault Diagnosis Protocols For Wireless Multi-hop Networks

A wireless multi-hop networks (WMN) is a self-organized, infrastructure-less net-work of mobile devices that are connected by wireless links. It has been widely used inmany fields. Because of the constrained resource, hazardous and complex environment,and dynamic topology structure, devices in a WMN are prone to failure which reducesthe reliability of the network. Fault diagnosis protocols are deployed to detect and locatefaulty devices, and furthermore support the followed fault isolation and fault recoveryoperations. Current researches on diagnosis protocol design in WMNs are done in an ad-hoc manner. Protocols are always designed based on general and conceptual models suchthat they are too general to be used in real applications and cannot handle requirementchanges well. This dissertation investigates a domain-oriented design methodology fordiagnosis protocols in WMNs, targeting job shop applications in the large-scale discretemanufacturing field. The main contributions are summarized as follows:1. It proposes a domain-oriented fault diagnosis protocol design architecture DPArch,which consists of three parts: the domain modeling part, the protocol design part,and the protocol simulation and verification toolkit part. Models are classified bythe constraint degree which is the key characteristic that inferences the design ofdiagnosis protocols. For now, there are three categories: static model, random dy-namic model and constraint dynamic model. The diagnosis protocols are designedaccording to the classification.2. For static models, based on a probabilistic fault model it proposes a proactive, test-based diagnosis protocol DPSM. The diagnosis process is limited in a single nodeand its neighbors such that the communication cost is reduced. The local and globalperformance analysis of DPSM are given. The experiment results show that theprotocol can correctly identify more than97%nodes when the percentage of fault-free units is about50%.3. Based on DPSM, a fault diagnosis protocol DPRDM for random dynamic modelsis proposed. The accuracy and consistency in a mobility environment are improvedby adjusting the strategies for the determination of device status and the transmis-sion of diagnosis information. The correctness proof of DPRDM under traditional fault model is given as well. The results of experimental analysis show that nodeidentification rate of DPRDM is higher than90%when the network connectivity ismore than70%and fault-free units percentage is50%.4. For constraint dynamic models, it proposes a diagnosis protocol DPCDM on thebasis of DPRDM. Utilizing the characteristics of constraint dynamic models, thenetwork connectivity is increased by dynamical power adjusting, which leads toan improvement of the protocol performance. A genetic algorithm is deployedto compute the optimized power configuration scheme. Experiment results showthat, compared to the average power configuration scheme, connectivity and nodeidentification rate are improved by15.7%and11.2%respectively.5. Based on DPArch, a prototype of WMNs fault diagnosis protocol design and anal-ysis toolkit (WPDA) is implemented. WPDA can be used to design, simulate, an-alyze and optimize diagnosis protocols for a specific domain. An evaluation indexsystem for diagnosis protocols is given as well. A case study of WPDA is presentedand the comparison shows that the performance of our protocol is better accordingto the evaluation index system.