【作者】 张勇；

【导师】 孟庆浩；

【作者基本信息】 天津大学 ， 检测技术与自动化装置， 2012， 博士

【摘要】 无线传感器网络（WSN）通常由大量小体积、低成本且环境适应性强的传感器节点组成，其集传感器技术、信息处理技术和网络通信技术于一体，可实现信息的采集、融合和传输。生化气体泄漏源检测与定位是WSN在环境监测领域的一个典型应用，其目的是寻找有效且精确的定位方法，及时地对污染源做出预警与定位。研究成果可广泛应用于有毒有害气体泄漏源查找与定位、危险环境监控、火源探测与预警等场合。本文对基于WSN的化学源分布式检测与定位方法进行了研究，重点开展了如下工作：第一、提出了传感网络生化气体泄漏源检测与定位理论框架，包括气体扩散模型构建和传感网络分布式信息处理两个方面。第二、提出了一种序贯分布式最小均方误差（D-MMSE）估计算法实现气体泄漏源检测与定位。推导了D-MMSE估计量及其均方误差表达式；构建了包含节点间信息增益和通信链路能耗两方面参数的信息融合目标函数，并对其求极值以完成路由节点选择；所选节点在其测量值和前一个节点估计结果基础上与邻居节点信息交互，实现了气体泄漏源位置参数估计量及均方误差的更新与传递。为降低通信链路能耗，按照估计量均方误差对邻居节点集的选择半径进行动态调整。通过计算机仿真验证了所提D-MMSE算法的性能。第三、提出了一种基于WSN的分布式序贯卡尔曼滤波方法用于实现生化气体泄漏源检测与定位。鉴于环境中气体扩散模型的高度非线性、非高斯特征，其核心算法由序贯扩展卡尔曼滤波算法（S-EKF）和序贯无迹卡尔曼滤波(S-UKF)算法组成。并通过仿真实验对这两种算法的性能进行了论证分析。第四、提出了一种基于协作多输入多输出（MIMO）分簇传感网络的生化气体泄漏源分布式参数估计方法。首先在簇内采用并行粒子滤波算法实现气体泄漏源状态参数的分布式估计，并得到估计量及其估计方差，然后根据估计量的方差以及方差的迹进行下一个簇集调度，最后在能量约束条件下运用凸优化算法实现了簇与簇之间的协作MIMO信息传输。更多还原

【Abstract】 Wireless sensor network (WSN) is usually composed of many small low-costsensor nodes with high adaptability to environments, which integrates sensortechnology, information processing and network communication technology to fulfillthe information collection, fusion and transmission. The detection and localization ofgas-leakage sources based on WSN is a typical application in the field ofenvironmental monitoring, which is aimed at finding an effective and accuratepositioning method to locate the source of pollution and give timely warning. Theresearch results can be widely applied to searching and positioning of toxic andhazardous gas leak sources, dangerous environmental monitoring, detection and earlywarning of fire sources and other occasions.This dissertation mainly focuses on the gas-leakage source detection andlocalization using distributed sensor networks. The main contributions can beconcluded as follows:Firstly, the theoretical framework for WSN based biochemical gas-leakagesource detection and localization is proposed. The proposed framework includes twoaspects, one is gas diffusion modeling, and the other is distributed informationprocessing.Secondly, a gas-leakage source detection and localization algorithm based ondistributed minimum mean squared error (D-MMSE) sequential estimation isproposed. In the proposed algorithm, the expression of the D-MMSE estimator and itscorresponding mean square error is derived; An information fusion objective function(IFOF) which combines the information utility measure and the communication costbetween sensor nodes is constructed, and the sensor node scheduling scheme isdesigned by optimizing the IFOF; For each selected sensor node, the estimator and thecorresponding mean square error are updated with its own observation and the noisecorrupted decision from the previous node, and the updated results are transmitted tothe next selected node by collaborating information within its neighborhood; Todecrease the energy consumption, the neighborhood radius is adjusted dynamicallybased on the mean square error. The performance of the proposed D-MMSE algorithmis verified through computer simulations.Thirdly, a sequential Kalman filtering theory is used for the biochemicalgas-leakage source detection and localization in the distributed sensor networks. Given the highly non-linear and non-Gaussian physical model of gas distribution inthe environment, the sequential extended Kalman filter (S-EKF) and sequentialunscented Kalman filter (S-UKF) are selected to localize the chemical source basedon the concentration detected using wireless sensor nodes. Simulation results validateboth algorithms.Fourthly, a distributed parameter estimation method for biochemical gas-leakagesources based on the cooperation of multiple inputs and multiple outputs (MIMO)cluster sensing networks is proposed. At first, the distributed state parameters of thegas-leakage source are estimated by the parallel particle filtering algorithm in thecluster, and the parameters’ estimators and variances are calculated. Then, the sensornodes scheduling in different clusters are implemented with the variances andcorresponding traces. At last, the information transmission with the collaborationMIMO method between two clusters is realized using convex optimization algorithmwith the energy constraint conditions.更多还原