Investigation On Geographic Routing And Information Fusion Of Wireless Sensor Networks With Complex Network Theory

A wireless sensor network (WSN) is a multi-hop self-organized network system that consists of a large amount of low-cost and small-size sensor nodes with limited sensing, computing and communication capabilities. Among various theoretical and application-oriented researches, the routing protocols and information fusion algorithms are fundamental for the data transmission and information processing in WSNs. The limited communication and computation capability and the limited energy resources of the sensor nodes bring the serious challenges for the design of the routing protocols and the information fusion algorithms.This dissertation discusses the geographic routing and information fusion problems in WSNs from the viewpoints of the complex network theory. With the small world network models and the consensus problems, the design and optimization of the geographic routing protocols and the distributed information fusion algorithms are thoroughly studied. The main contributions are summarized as follows:1. Geographic routing protocol based on the small world modelsGeographic routing is a kind of popular routing techniques in WSNs. The presence of void area often greatly affects the greedy forwarding. With the navigability of the small world networks, this dissertation develops a topological awareness-oriented geographic routing algorithm. The algorithm first defines the concept of topological awareness which contains topological information of non-neighbor nodes. Then a logical network having properties of the small world network is constructed with the topological awareness. Finally, the information transmission is conducted through the navigability of the small world network and thus the probability of entering the void area is decreased. The topological awareness is the key to the development of the routing algorithm. After defining the effective topological awareness and analyzing its structure characteristics, a generating method to construct the effective topological awareness is introduced with the small world network. A number of simulations are conducted and the results show that the proposed routing algorithm can greatly enhance the packet delivery ratio without the need of the recovery strategies, and experience nearly shortest forwarding path, especially in the network seriously obstructed by obstacles. As a result, the energy consumption during the data transmission is decreased, and the lifetime of the network is prolonged.2. Distributed estimation fusion techniques based on the consensus algorithms The consensus-based distributed estimation fusion techniques require multiple information exchanges and state updates to reach the consensus state, which is the global optimal estimate. These operations not only increase the energy consumption of the sensor nodes but also prolong the consensus convergence. This dissertation combines the consensus-based estimation techniques with signal processing methods to improve the convergence performances. The state predictors and adaptive filters are adopted. The state predictors perform the instant state estimation, and thus can flip over the redundant state of convergence process. The adaptive filters can adaptively update the weighting matrix and improve the convergence. The simulations results demonstrate that the proposed methods indeed accelerate the convergence rate of the node state estimates, and thus reduce the node energy consumption.3. Distributed tracking algorithms based on the consensus algorithmsThe design of the distributed Kalman filters and the distributed Sigma-point Kalman filters is considered for the application of the consensus-based distributed tracking in WSNs. The idea is to enhance the consensus performance of the different node states by combing the consensus algorithms with linear Kalman tracking techniques. For the nonlinear dynamical system, the Sigma-point based weighted statistical linear regression is utilized to linearize the state equation and the measurement equation. The simulation results demonstrate that the proposed algorithms not only improve the accuracy of the state estimate, but also largely enhance the consistency of the state estimate at each node. The simulation results also reveal that it is fundamental to improve the consistency of the different state estimates for the performance enhancement.