Research On Wireless Location Algorithm To Mitigate The Effect Of NLOS

In the past few decades, wireless location technology which is used to estimate the location coordinate of mobile station (MS), has gained wide attention. Especially, the location based service (LBS) is widely used. Wireless location is usually divided into two steps, it firstly obtains the required parameters for location such as time of arrival (TOA), time difference of arrival (TDOA), angle of arrival (AOA) as well as Received Signal Strength (RSS); then, using the obtained parameters and the corresponding algorithm to achieve estimated location coordinate of MS. This paper investigates the second step of wireless location and researches the wireless location algorithm based on TOA.Non-line-of-sight (NLOS) propagation is the main factors to affect the accuracy of the wireless location algorithm. In the line of sight (LOS) environment, the linear least squares algorithm is a low complexity algorithm, by choosing a reference point to transform the nonlinear equation relationship to linear relationship, and then using the least square algorithm to achieve the estimated location coordinates of MS. In the latest literature, the minimum measurement distance is selected as the reference point. By directly analyzing the objective function of the least square algorithm, it is proposed to select the smallest residuals as a reference point. The proposed algorithm is superior to the minimum measurement distance as a reference point. In NLOS environment, combining the proposed LOS algorithm with the existence of NLOS algorithm (the residual weighting algorithm and three-step algorithm) to form new algorithms, the simulation results demonstrate the new algorithms have better performance than the existing algorithms.In NLOS environment, quadratic programming algorithm can significantly reduce the impact of NLOS error and does not need any prior information about the NLOS error. But the initial step of this algorithm adopts the maximum likelihood (ML) algorithm to obtain distance estimation to substitute the real distance of the error covariance matrix. The ML algorithm is adaptable to LOS environment, it is found that the measurement distance is more suitable than the initial estimated distance obtained by ML algorithm. In addition, by considering the geometric relationship between the MS and fixed station (FT), data processing may reduce the impact of big NLOS error. Again, the iterative method can be introduced to further improve the performance of the quadratic programming algorithm. Based on the above three points, this paper proposed the improved quadratic programming algorithm, simulation shows that the improved algorithm significantly improves the performance.In NLOS environment, range scale algorithm is proposed in2004, its main idea is to introduce three new variables to establish nonlinear programming, and the estimated location coordinates are obtained. But, it is high computational complexity and only suitable for three FTs. For three FTs, I introduce a new variable to replace the squared term, putting the nonlinear objective function of the nonlinear programming into linear form, reducing the computational complexity significantly. In addition, the simulation results show the location performance is greatly improved. Moreover, I extend the range scale algorithm to adapt more than three FTs.In NLOS environments, a method for a high location accuracy is first to identify NLOS FTs, and then using the LOS FTs to achieve the estimated location coordinates of MS. When the NLOS error is a Gaussian distribution, by making the total error probability (The sum of the probability of false alarm and the probability of missed detection) minimum, the analytic expression of the optimal single threshold is derived. The simulation results show the optimized single threshold is superior to the traditional NLOS identification method. In addition, if the priori probabilities of the LOS and NLOS are known, I proposed the novel double threshold NLOS identification method. The simulation results demonstrate that the double threshold schemes can further improved the performance of NLOS identification.