Researches On Indoor Localization Technique Of Wireless Sensor Network Based On Off-grid DOA Estimation

Recently, with the rapid development of wireless network Smart Grid and Internet of Things have emerged frequently in our daily life. Wireless terminals are ubiquitous to increase the requirements of localization services with wireless network. Wireless sensor network (WSN)-which is a new wireless technique with the advantages of low cost, low energy consumption, flexible structure, and so on-attracts a widely attention both in the industrial and science research field. The mature researches on network structure and protocol make the application of WSN in the indoor localization system have huge potential. The localization service for indoor environment has get atten-tion in smart building, health care, modern agriculture, warehousing management, and so on. In WSN localization system, except for the very important wireless technique the localization algorithm is also important. Currently, many localization algorithm-s can effectively localize the indoor targets. However, complex indoor environment decreases the localization accuracy and requires the researches on the algorithm im-provements. This thesis is about localization technique in the indoor WSN. Direction of arrival (DOA) estimation method is a localization algorithm with high accuracy based on the array signal processing, which has been applied in the indoor localization just now. The application is lack of mature theoretical understanding and quantitative analysis and faced with a lot of challenges. A few years ago, a breakthrough theory of signal processing, named as compressed sensing (CS), broke the constraint of Nyquist frequency on signal compression and recovery, which also brings interesting opportu-nities to DOA estimation and increases the probability of DOA application in indoor system. The CS theory decreases the requirement of array size for DOA estimation, brings about higher estimation accuracy and smaller resolution.Of course, the existing CS based DOA estimation methods are faced with some problems and challenges in the indoor applications. The problem of grid mismatch attracts the most attention, which is brought about by the inevitable quantization error in the discrete division of space domain so as to limit the estimation actuary. In addition, I focus on three key problems in the indoor environment in this thesis, which are:· The phenomenon of aliasing and ambiguity exists during the DOA estimation for coherent sources in the application of indoor localization, which causes the decreasing algorithm stability.· In a real scenario multiple arrays may estimate the DOA information of same sources, the DOA estimation of this common source with different arrays should cause repeated computations. Thus, we need more sensors in a array to localize a known amount of targets which will increase the hardware cost.· We can not neglect the packet losses during the data transmission in indoor WSN, which causes the increasing estimation error. The existing solutions can effec-tively deal with this problem, but they require the improvements of hardware and algorithm in sensors to increase their energy consumptions.In this thesis, I carry out my research work from the above three points, which include:·The super resolution technique for the problem of gird mismatch is applied in the DOA estimation algorithm to form a off grid (OG) DOA estimation method with high DOA estimation accuracy, where continuous atomic basis and atomic norm minimization are used to overcome the quantization error of grid brought about by the discrete division of space domain. An exactly accurate off-grid algorithm with decoherence processing is proposed for the DOA estimation of coherent sources. The structure of reduced signal sub-space for the covariance matrix of coherent sources is deeply analyzed. The signal vector including the complete DOA information of sources is constructed. The atomic basis in the continuous spatial domain and sparse spatial model are built up. The minimization problem of l1 norm in the original discrete domain is transferred into that of total variance norm, which can be solved via semidefinite programming (SDP).· In the real scenarios, we use a joint DOA estimation algorithm to solve the problem of DOA estimation with multiple arrays, which means all arrays syn-chronously estimate the DOA information via data sharing. Two situations, in-cluding distributed sparse arrays in the spatial domain and timeslot array in the time domain, are discussed. To solve the problem of distributed sparse arrays, I build up a joint spatial sparse (JSS) model for the common and innovated sources localized by the arrays. A visual uniform linear array (ULA) is modeled for each sparse array so that the practical observed signal can be seen as the compressed data of the signal observed by the visual ULA. Combinative atomic (CA) nor-m is used to describe the JSS model and transfer the DOA estimation problem into a minimization of CA norm which can be solved by SDP. Similarly, in the multiple sources localization a dynamic scheme is proposed to use the samples of random sensors during multiple timeslots to compliment DOA estimation. Timeslot array is modeled and the problem of DOA estimation is transferred in-to the minimization of atomic norm and solved by SDP.· The reliability of data transmission in the indoor WSN is studied in detail, and the impacts of link quality and network protocol are focused on. We design a scene-oriented WSN simulation and optimization platform to realize the perfor-mance evaluation of different practical scenarios. Considering that the existing improved methods always have the constraint on energy consumption of sensors, a post-processing data transmission method is proposed. The packet losses dur-ing wireless data transmission are seen as the random measurement processing in CS theory. The data can be recovered via the joint optimization of l1 and l2 norms. This method also has the function of denoising. The algorithm can be only implemented in the sink node with abundant resources, which requires no cost of the energy limited sensors to promote performance. Finally, by combin-ing these three methods for different problems I propose an indoor distributed localization technique. Multiple sensor nodes are used to form a network array to implement DOA estimation. The locations of sources are obtained via dis-tributed ADMM algorithm in the base station in the network.I compare several methods to evaluate the performance by using simulation tests to prove the validity of the proposed methods in the thesis from the indicators about es-timation error, estimation success probability, estimation resolution, and so on. All the proposed methods provide a new available scheme for the theoretical research on indoor localization system.