Research On Optimizing Parameter Grid For Direct Position Determination Of Emitters

The performance of traditional two-step positioning approach degrades when it is applied in complex scenarios(such as indoor scenario and scenario with finite observation resources).In order to improve the sub-optimal positioning performance,researchers propose a novel and optimal positioning approach-the Direct Position Determination(DPD)approach.The DPD approach jointly processes all measurments and directly es-timates the target position from the raw measurments,which avoids the information loss caused by the step-by-step processing in two-step positioning approach.Consequently,the DPD approach has drawn much attention and can significantly promote the position-ing performance in complex applications.As a new and developing approach,the DPD has its own problems to be solved.For example,the multi--target DPD problem is a big challenge because of the curse of dimensionality and the high computational complexity of exhaustive searching is intractable.Besides,meshing the parameter space results in the model quantization error inherently.Regarding the problems mentioned above,in this dissertation,the DPD approaches for emitters in different scenarios and its application to the indoor communication system are studied from the view of optimizing the parameter grid.The main contributions of the dissertation are summarized as follows:(1)With the purpose of jointly estimating the order and parameters of the DPD model and eliminating the quantization error,the indoor DPD model is built using the dynamic position grid and under the framework of sparse recovery a sparse Bayesian learning based indoor DPD algorithm is proposed,by combining the EM algorithm and the GD algorithm.Compared with the exsisting fixed grid based DPD algorithms,the proposed algorithm is more robust to the multipath interference and has better positioning performance.(2)In order to jointly solve the multi-target DPD problem and the ”intermittent emis-sion-sources” association problem,the emitter positions are regarded as the associa-tion labels and the emission statuses are estimated using the K-means clustering algorithm.A MVDR based DPD algorithm for multiple intermittent emitters is proposed,by decou-pling the emission statuses and the emitter positions.By comparison with the exsisting alternative projection based DPD algorithm,we show that the proposed algorithm does not rely on the prior information about the number of emitters and the initial position estimations.(3)To ruduce the computational complexity of the exhaustive searching,based on the Pincus theory the multi--target DPD problem is transformed into multiple single--target position expectation problems.According to the importance sampling technique,the exhaustive searching is replaced with parameter sampling and an importance sampling based maximum likelihood DPD algorithm is proposed.As a result,the desired position expectations are estimated using the parameter realizations and the inherent grid quantization error is eliminated.Compared with the exsisting DPD algorithms,the proposed algorithm has better performance with finite observation resources.(4)A FFT based TDOA DPD algorithm and its application to the indoor commu-nication system are studied.The computational load of the TDOA position spectrum is decreased by two order of magnitudes.Compared to the two-step positioning approach,the proposed algorithm has better performance with cost of increasing few computational load.The mentioned DPD algorithms and implementation scheme are verified to be ef-fective using numerical simulations or real data collected from the indoor communication system.