Gas-leakage Source Detection And Localization Based On Distributed Estimation

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.