Research On Node Self-Localization And Development Platform In Wireless Sensor Network

Wireless Sensor Network (WSN) integrates technologies on sensor, nested computation andwireless communication. It has the capacity to cooperatively sense, collect and processinformation within covered areas, and then transmits them to interested observers throughself-organized multi-hop network. In doing so, it connects the physical world of information withthe logic world of information, greatly expanding the scope of current network and vastlyenhancing mankind's ability to understand the nature. Wireless sensor networks can be set upwithout the support of infrastructure in a quick and flexible manner. Its extensive and significantpotential applications in a variety of fields make it an emerging research hotspot in informationscience.This thesis studies the node self-localization technology in WSN and constructs a WSNdevelopment platform. The former is one of the key technologies in WSN. The latter can be usedto practically evaluate and validate algorithms' performances. The main contents of this thesis canbe listed as follows:The Simulated Annealing based localization algorithm (SAL) and the Evolution Strategybased localization algorithm (LESS) are investigated. Both algorithms belong to the catalogue ofintelligent algorithm-based centralized localization algorithm. By defining the node-localization asan optimization problem, the Simulated Annealing algorithm and Evolution Strategy algorithm areput into use. The structures and procedures of SAL and LESS are introduced and analysed. Theissue of how to set parameters for the algorithms is also covered. By varying scenario settings,multiple simulations are run to evaluate the two algorithms' performances in depth. Theirapplication conditions are obtained.A novel Ant System based node localization algorithm is proposed. By discretizing anon-anchor node's estimated coordinates to a limited number of choices in each step, the nodelocalization procedure is built as a series of combinatorial optimization processes. Therefore it isfeasible to use Ant System in node localization. Based on the feasibility analysis, the Ant Systembased localization algorithm (ASL) is presented. It uses a periodically restarting Ant System todrive non-anchor nodes' estimated coordinates towards their actual positions. The relationshipbetween the algorithm's parameters is discussed. The setting of key parameters is investigated. By varying scenario settings, multiple simulations are conducted to validate the practicality of ASLand to examine its performance. Based on the results of simulations, reference values of networkparameters are given to facilitate the setting-up of a WSN. These values are aimed at achieving agood balance between localization accuracy and networking cost, including energy consumptionand deployment expenses.This thesis also elaborates the process of building a practical and easy-to-use developmentplatform to satisfy the critical need for experimentation in the research and development of WSN.TinyOS is selected as the software sub-system of the development platform after analyzingTinyOS and the software sub-system of GAINZ. Based on comparison between the GAINZ nodeand MICAz node, modification is done to TinyOS so that it can run correctly on a GAINZ node.The original version of SNAMP, a WSN analysis and monitoring program, does not providesufficient functions. To meet the needs of development, three new function modules are added toSNAMP. By combining GAINZ node, the modified TinyOS and the expanded SNAMP together,a new and completely open-source WSN development platform is established. For the purpose ofresearch as well as demonstration, revison is made to the Surge application in TinyOS which isused to set up a tree network and converge data to a sink node. The revised Surge application canform a mesh network which is more robust and has more balanced energy consumption. It is thenrun on the newly-built development platform. The results presented in SNAMP validate thepracticality of the platform.