Research On Key Technologies For Node Localization In Wireless Sensor Network

Wireless sensor network (WSN) can be applied to many fields that related to the national economy and people's livelihood, and node localization is the prerequisite for WSN to carry out certain applications. Due to the high precision ranging ability and the low implementation complexity, impulse radio ultra-wideband (IR-UWB) is now regarded as the most appropriate physical layer technology for WSN. In the IR-UWB based WSNs, to utilize the fine time resolution of IR-UWB most, node localization is realized based on TOA estimation based ranging.Researches on the key technologies of node localization for IR-UWB wireless sensor network are taken in this dissertation. Based on analysis of the research status quo at home and abroad, problems and defects of the existent researches on localization system are pointed out. These problems and defects exist both in the internal mechanisms of the three modules of localization system (i.e. the ranging module, NLOS processing module and the positioning module), and in the external communications between the modules. The purpose of this dissertation is just to carry out some mends as well as improvements on these problems and defects, so that the theoretic system of ranging and positioning for IR-UWB wireless sensor network can be consummated to certain extent. Principal contents of this dissertation are summarized as follows.Firstly, researches on the TOA estimation algorithms for the ranging module are carried out. Four different TOA estimation algorithms that suit for different situations are proposed, i.e., I-CLEAN-GML, ED-TC-MMR, Two-Step and MF-TC-JM. The I-CLEAN-GML algorithm is an improved version from the CLEAN-GML algorithm, whose implementation refers to iterative correlation detections and estimations, while the implementation process has been greatly optimized in I-CLEAN-GML. This algorithm can mine the precise ranging ability of IR-UWB to the most, thus it can be adopted in the future WSN applications of which the node hardware has developed to a high enough level, while now its results can be referred as the performance benchmark for researches on the other TOA estimation algorithms. ED-TC-MMR is an energy detection based non-coherent TOA estimation algorithm, where the normalized detec- tion threshold can be optimally set based on the MMR value that extracted from the received energy samples. This algorithm outperforms the other existent non-coherent TOA estimation algorithms at nearly all SNR ranges, thus it can be adopted in current WSN applications where the node hardware level is not so high and requirement on ranging precision is not so strict. The Two-Step algorithm jointly employs energy detection and correlation detection. The energy block that contains DP is detected by energy detection in the first step, and then the precise location of DP within that detected block is obtained by correlation detection in the second step. The performance of Two-Step is better than ED-TC-MMR, while the computation complexity is not so high. Thus it can be adopted in those WSN applications where the node hardware ability is high enough and the requirement on ranging precision is higher than what ED-TC-MMR can provide. MF-TC-JM directly detect the location of DP from the match-filtering output of the received signal, where the detection threshold can be optimally set based on the a joint metric in terms of kurtosis and root mean square delay spread of the match-filtering output. The computation complexity of MF-TC-JM is a little higher than Two-Step, while the ranging performance has been relatively improved, i.e., the trade-off strategy of MF-TC-JM favors the ranging precision to some extent. MF-TC-JM can along with Two-Step provides ranging schemes for WSN applications that seek for trade-off ranging strategies. As for ED-TC-MMR, we even plant it in the IR-UWB testbed that is developed by Harbin Institute of Technology Shen Zhen Graduate School. The validity of ED-TC-MMR can be further proved by real ranging experiments, and also the ranging ability of the testbed with current configuration can also be investigated.Secondly, effects of the source end parameter settings on ranging and reliability of the output results of ranging module are studied. The source end parameters refer to the transmitter-related parameters, including the pulse shape, pulse bandwidth and pulse repetition interval. Effects of the settings of every parameter on ranging are first qualitatively predicted through theoretic analysis, and then a large amount of simulations under IEEE 802.15.4a channel model are conducted to obtain quantitative results. Both coherent ranging and non-coherent ranging are considered in the simulations to make the results more conclusive. The purpose of evaluating the reliability of the output TOA estimation results is to provide the reliability information for the positioning module so that the ultimate positioning precision can be further im- proved. The method of reliability evaluation adopted in this dissertation is to classify the TOA estimation results into different reliability levels according to the threshold setting metric, and the probability density functions of the TOA estimation errors at each level are modeled.Thirdly, researches are conducted on how to handle the NLOS problem. NLOS effect is one of the main factors that affect the ranging precision under dense multipath environments. In this dissertation, researches on NLOS processing are carried out through two ideas, one is to identify the channel state (is NLOS channel or not), and the other is to estimate the magnitude of the NLOS error. The proposed NLOS channel identification method do not require accurate channel estimation; it directly extract the product of DP-to-average and peak-to-average from the received signal segment samples as the identification metric, and likelihood ratio test on the metric is conducted to identify the channel state. The proposed NLOS error estimation method is deduced from the signal propagation path loss model, and two single paths, i.e. DP and MCP, have to be detected from the received signal to implement the method. To compare NLOS channel identification and NLOS error estimation, the former is easier to implement, while the latter is more straight and efficient. In practice, selection of the NLOS processing ideas should be dependent on the concrete situations.Fourthly, the single node positioning algorithm and the network cooperative positioning method for the positioning module are studied. Single node positioning is the base of network positioning. In this dissertation, an integrative ranging-positioning platform is built, where the range estimation results used by the positioning algorithms are obtained from the processing of ranging module on the received ranging signal (not from randomly generation according to certain probability distribution models), and in addition, incorporation of the reliability information provided by the ranging module and the NLOS state information provided by the NLOS processing module is also considered. Three different positioning algorithms which differently utilize the reliability information and the NLOS state information are proposed, including LS, WLS, and ML. The advantage of network cooperative positioning over single node positioning is that the positioning coverage can be enlarged and the NLOS effects can be greatly mitigated. In this dissertation, an improved distributed network cooperative positioning method (i.e. I-AHLos) based on the existent AHLos is proposed. In I-AHLos, the selective node upgrade strategy and location weight strategy are intro- duced, thus the successive improvement of positioning accuracy is achieved, and the problem of cumulative error propagation is solved to some extent.