Advanced Science and echnology Letters Vol.121 (AS 2016), pp.344-349 http://dx.doi.org/10.14257/astl.2016. Indoor Location Algorithm Based on Kalman Filter Zhang Ya-qiong 1,*, Li Zhao-xing 2, Li Xin 3,*, Lv Zhihan-han 4 1 Information Engineering school of Yulin University in Yuyang, Yulin, Shaanxi Province, China 2 Information Engineering school of Yulin University in Yuyang, Yulin, Shaanxi Province, China 3 School of Urban Design, Wuhan University, Wuhan, China 4 SIA, Chinese Academy of Science, Shenzhen, China * Corresponding Authors Abstract. For onboard single-station passive direction-finding and location, if there is any abnormal error in the observation data, the extended Kalman filter (EKF) algorithm adopted thereby will cause inaccurate location result. In order to improve algorithm robustness, the robust equivalent gain matrix is constructed according to the standardized prediction residual error and the robust EKF algorithm is applied to the onboard single-station passive directionfinding and location. In allusion to the low efficiency of the robust EKF algorithm, the single-station passive location algorithm based on the improved extended Kalman filter is proposed in this article on the basis of combining F distribution statistic, and meanwhile single abnormal error and continuous abnormal error are added in the observation value to test the algorithm resistance to different abnormal errors. he simulation shows that the algorithm proposed in this article can well weaen the influence of abnormal errors on position estimation and the algorithm based on F distribution discriminant can improve location efficiency. Keywords: Indoor location; Robust EKF filter; F distribution discriminant; Location accuracy 1 Introduction In allusion to the low efficiency of the robust EKF algorithm, the single-station passive location algorithm based on the improved extended Kalman filter is proposed in this article on the basis of combining F distribution statistic, and meanwhile single abnormal error and continuous abnormal error are added in the observation value to test the algorithm resistance to different abnormal errors. he simulation shows that the algorithm proposed in this article can well weaen the influence of abnormal errors on position estimation and the algorithm based on F distribution discriminant can improve location efficiency. ISSN: 2287-1233 ASL Copyright 2016 SERSC
Advanced Science and echnology Letters Vol.121 (AS 2016) 2 Onboard Single-Station Passive Direction-Finding and Location Model For convenient calculation, the two-dimensional onboard single-station passive direction-finding and location model is taen as an example for relevant analysis, the state equation and the observation equation of the location model are also established, and meanwhile the observation equation is linearized. As shown in Fig.1, the target radiation source is located at xt, y t. An airplane is assumed to start from the origin and mae uniform linear motion at speed v and meanwhile change motion direction, wherein the coordinate of the th observation point of the airplane is x, y, and is the azimuth angle measured at the th observation point. y x, y t t o 1 x, y x Fig. 1. Location model If the airplane is taen as the reference system, then the state equation of the target radiation source is as follows: 1 X X W (1) herein: W N 0, Q,,, X x x y y xt x, vx, yt y, v y Copyright 2016 SERSC 345
Advanced Science and echnology Letters Vol.121 (AS 2016) 1 0 0 0 1 0 0 0 0 1 0 0 0 1 2 2 0 0 0 2 0 2 he observation equation of the onboard single-station passive direction-finding and location can be expressed by the following formula:, Z h X V (2) In the above formula, V N 0, R,,, Z 1 2 n yt y X xt x, vx, yt y, v y h, X arctan xt x When the airplane is taen as the reference system, the location model can be converted into the problem that the fixed single-station tracs the target radiation source under uniform motion, and Kalman filter algorithm can be adopted for calculation. he motion model can be converted as follows: 1 X X W (3) Z yt y arctan V xt x (4) 346 Copyright 2016 SERSC
Advanced Science and echnology Letters Vol.121 (AS 2016) According to the model, the state equation is linear while the observation equation is nonlinear, and the linearized observation matrix is as follows: Z H X Z Z Z Z,,, x x y y 2 2 yt y xt x,0,,0 2 2 2 2 xt x yt y xt x yt y (5) he extended Kalman filter algorithm can realize the linear approximation of the nonlinear system to improve location accuracy. For a given model, the extended Kalman filter therefore is further predicted as follows: 1 ˆ X X (6) 1 1, 1 Z h X (7) he prediction covariance matrix is as follows: 1 P P Q (8) he filter estimation and the corresponding covariance matrix thereof are as follows: 1 1 ˆ 1 Xˆ X 1 1 ˆ 1 K Z Z 1 1 1 1 P 1 P I K H (9) (10) Copyright 2016 SERSC 347
Advanced Science and echnology Letters Vol.121 (AS 2016) he observation estimation is as follows: 1 1, ˆ 1 1 Zˆ h X (11) EKF gain matrix is as follows: 1 1 1 1 K P H H 1 1 1 P H R 1 (12) For a given nonlinear location model, EKF algorithm can effectively solve nonlinear problem to obtain a relatively optimal state estimation. 3 Conclusion In this article, the robust EKF algorithm is applied to the onboard single-station passive location, and meanwhile the robust EKF algorithm based on F distribution discriminant is also proposed. he simulation shows that this algorithm can resist to the influence of abnormal error and has good location performance. When significant abnormality exists between the state equation and the onboard motion curve and both the observation equation and the state equation are influenced by abnormal error, the passive location estimation will also be significantly influenced and accordingly have deviation. In future, these complex environments will be considered in order to further improve algorithm robustness. Acnowledgments. his wor was supported by the Shaanxi Provincial Science and echnology Department of Agricultural science and technology innovation and research project(2015ny047), Yulin Municipal Science and echnology Bureau of research projects(2014cxy-3-03), National Natural Science Foundation of China (No. 51408442). Reference 1. Wang, K.: Overcoming Hadoop Scaling Limitations through Distributed as Execution 2. Zhang, S., Zhang, X., Ou, X., After we new it: empirical study and modeling of costeffectiveness of exploiting prevalent nown vulnerabilities across iaas cloud. Proceedings of the 9th ACM symposium on Information, computer and communications security. ACM, (2014) 3. Bao, G., Mi, L., Geng, Y., Pahlavan, K.: A computer vision based speed estimation technique for localizing the wireless capsule endoscope inside small intestine, 36th Annual 348 Copyright 2016 SERSC
Advanced Science and echnology Letters Vol.121 (AS 2016) International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), August (2014) 4. Gu, W., Lv, Z., Hao, M., Change detection method for remote sensing images based on an improved Marov random field. Multimedia ools and Applications. (2016) 5. Lu, Z., Esteve, C., Chirivella, J., Gagliardo, P.: A Game Based Assistive ool for Rehabilitation of Dysphonic Patients. 3rd International Worshop on Virtual and Augmented Assistive echnology (VAA) at IEEE Virtual Reality 2015 (VR2015), Arles, France, IEEE, (2015) Copyright 2016 SERSC 349