Research Of Indoor Locating And Tracking Technology Based On Rssi Distance Measurement

With the development of communications and networking technologies, and the growing demand for indoor position information services,which made the indoor location and tracking technology become an important field of current scientific study. Because of the simple signal detection equipment, low-difficult detection mechanism and low-cost property, the localization technology based on the received signal strength indication(RSSI) has became a mainstream method of indoor location and tracking technology.The propagation loss model of indoor signal is the key to the indoor location and tracking technology based on RSSI. But due to the complexity of the indoor environment, the multipath effects, the traditional propagation loss model with fixed-parameter has poor environmental adaptability, which lead to greater ranging and locating errors. Besides, the traditional method of neural network training for propagation loss model has shown disadvantages of more training samples and more workload of hardware. To overcome the problem mentioned above, under the variable density sampling mode, a method for propagation loss model construction based on the combination of grey theory with Radial Basis Function(RBF)neural network was put forward. Based on the grey theory, with more training samples forecasted with fewer samples and a part of the original samples, training samples were reconstructed for the RBF neural network to build the propagation loss model. The indoor propagation loss model can be built accurately with less training samples, which can meet the precision requirement of the indoor localization and reduce the workload of sample collection greatly.The indoor location and tracking technology based on geometrical method can be divided into two main categories: range-free and range-based. The range-free indoor location algorithms such as centroid method and weighted centroid method have low-requirements of hardware, but have large location errors. In the practical application, the range locating circle of the range-based algorithms such as trilateral location, the maximum likelihood, and least squares method, can not meet at a common point, which lead to be difficult to obtain the optimal solution, at the same time, the reference nodes different influences to the target position are usually not taking into account, which makes difficult to get high positioning accuracy To overcome the problem mentioned above, a Particle Filter(PF) algorithm with fitness optimization based on errors compensation was proposed in this paper, with this algorithm, the algorithm range location problem can be converted into an optimization problem of nonlinear equations, with considering the reference nodes different influences to the target position, the location deviation can be reduced to a certain extent, and the location result can be closer to the true position.In real-time location study, using positioning algorithm alone will lead to form an unreasonable, even wrong tracking path because of ignoring the prior information. Besides, because most of the reality tracked target paths are nonlinear and non derivable problem such as broken line and round trip, standard Kalman Filtering will be difficult to solve the forecasting problems of nonlinear system, and Extended Kalman Filtering(EKF) having some problems such as a high computational complexity caused by linearization for nonlinear systems, sensitivity to noise, and filter divergence caused by the error accumulation, which cause large forecast errors for target tracking. To overcome the problem mentioned above, a Unscented Kalman Filter(UKF) algorithm based on turning angle compensation was proposed in this paper, with this method, the nonlinear and non derivable path can be converted into a derivable linear or nonlinear path, and the location errors can be reduced by filter forecasting, which makes the precision of real-time positioning and tracking can be improved.The proposed algorithm in this paper were validated and improved by a way of combining theoretic analysis, computer simulation, and experimental study, which also provided a reference for the practical application of the algorithm.