Study On Indoor Positioning Algorithms Based On Multi-data Fusion

The Pedestrian Dead Reckoning(PDR)technology has been widely used in positioning and navigation systems,especially in complex scenes(indoor and underground)because of its unique advantages.It avoids the environment interference and can obtain high precision with low cost.However,absolute position information fails to be obtained in that the PDR can only achieve the relative positioning in the sensor coordinate system.What's more,the accumulative errors caused by the inertial sensors lead to lower accuracy drastically.Aiming at the above problems,a complete indoor positioning scheme based on multi-data fusion is presented,which fully exploits the physical environment characteristics such as geomagnetism,Wi-Fi,and building structures to correct errors and achieve real-time and efficient indoor positioning.In view of the large drift noise in PDR,two improvements are proposed in this thesis: Firstly,after the noise characteristics analysis of inertial sensors,a wavelet de-noising method based on multi-scale threshold function is proposed.The experimental results show that the Signal Noise Ratio(SNR)has been improved by 5d Bm compared with the tradition ones,and the relative navigation error within 100 meters is reduced by 2.69%.In addition,in order to improve the convergence rate of the orientation,the simplified factored quaternion algorithm(FQA)is utilized to optimize the initial angle instead of the original default value.The convergence time of is reduced from 20 s to 3s,and the performance is greatly improved.What's more,multi-source information fusion technology has been used in this thesis to improve accuracy.First,by learning the unique physical environment characteristics(geomagnetic and Wi-Fi signal)in the indoor environment,the special ‘points' are selected as reference landmarks,which can provide initial information and correct location errors in positioning.Subsequently,to enhance the adaptability to the spatial map,a matching algorithm is designed to better solve the problem of ‘cross wall' and improve the robustness of the system.Finally,a particle filter fusion algorithm based on multi-source information is proposed to estimate the pedestrian motion state so as to achieve accurate positioning.The experimental results show that compared with single PDR algorithm,the average error of the proposed algorithm decreases by an order of magnitude and is controlled at 0.0257 m.In addition,the average positioning accuracy is controlled within 2m(100m).