Research And Implementation Of Wi-Fi And Inertial Sensor Fusion Location Algorithm Based On Particle Filter

With the development of Intelligent terminal related technology, they provided opportunities to achieve urban canyons, indoor blind and closed underground space location service. Subsequently, variety of indoor positioning schemes based on wide area and local area arises at the historic moment. However, there are many restrictions on the indoor environment, positioning device, user state, localization terminal state and location area for the most indoor positioning scheme, it is urgent to research and adapt to the environmental change, stable, reliable indoor space navigation and positioning technology.At present, Wi-Fi network and intelligent mobile terminals have become popular and embedded of Wi-Fi module and variety of inertial sensors, so the intelligent terminal platform for research and development of indoor positioning system is feasible and practical. This paper takes indoor environment as the breakthrough point, through experiments find out factors influencing the effect of indoor positioning, using Wi-Fi positioning model and pedestrian dead reckoning model, are used to obtain absolute position information and relative position information, so as to design appropriate positioning algorithms and fusion scheme to resolve or weaken the influence of positioning. The main research contents are as follows:(1) The Wi-Fi signal characteristics analysis as the starting point, respectively, from two aspects of time and space. In these two aspects and four conditions(i.e., considering only the short time interval the influence of random factors and micro distance changes, can ignore the influence of random factors time interval and spatial distance changes) analysis of the Wi Fi signal characteristics.(2) Based on the KNN algorithm, combined with the characteristics of the Wi-Fi signal, considering the specific region and the actual indoor location environment, the adaptive Wi-Fi algorithm is proposed. Including the artificial vector domain clustering algorithm, in order to meet the needs of specific regions of the location service; based on a more stable Wi-Fi signal AP and positioning terminal between spatial position relationship of stable characteristics, improved the KNN algorithm, to weaken with time Wi-Fi signal fluctuations effect on the localization; put forward the recognition of monomeric malfunction AP method, enhances the Wi-Fi positioning of mutation factor identifying and eliminating ability; and put forward a rough identification fingerprint database of the effectiveness of the method using the corresponding relationship between Euclidean distance and positioning performance of Wi-Fi positioning.(3) The combination of different sports scene and pedestrian step statistics, analyzes the characteristics of inertial sensor signal. According to the analysis results, the digital FIR bandpass filter to preprocess the sensor signal; Through the number of particles in the sample by the pedestrian in stationary, normal walking and running the consumption of time and one step time correlation and adaptive peak detection algorithm to obtain sample particle number to identify the motion state; draw conclusions on the course, due to the different environment declination and different motion scenarios positioning terminal placed angle different, in the case of the coexistence of multiple motion scenes, the course information obtained through the inertial sensor cannot be applied position in the room.(4) Based on particle filter, using two kinds of positioning model of strong non similarity and complementarity, in the study, the combined positioning method based on Wi-Fi is given priority to, and the second is supplemented by the method of pedestrian dead reckoning. The importance distribution function is established from the point of view of easy realization, wide distribution and full use of the latest observation information. The local nonlinear filtering algorithm is adopted to the posterior probability distribution of the system state, such as the EKF for initial estimation, determine the domain of particle distribution, in this domain the importance distribution function of the Gauss distribution generated by particle filter. Finally, based on the particle filter to complete the positioning.