Research On Dynamic RF Map Establishment Method Based On RBF Neural Network

Currently, the indoor positioning technology has become a hot research topic in the field of wireless communications. In existing indoor positioning technology, RF-based fingerprint localization algorithm has been widely concerned and researched with universality and efficiency. RF fingerprinting positioning algorithm establish the position of the radio frequency fingerprint by receiving location-related signal strength values, on this basis, the positioning method matches the position. This method consists of off-line training stage and on-line locating stage. In the off-line training stage, the main purpose is to construct the radio map; in the on-line locating stage, according to the fingerprint map, we use a certain location algorithm to get the location information finally.However, the complexity of indoor environment and the variability of indoor wireless signal will bring great difficulties to establish the radio map, specifically in the training workload, the poor environmental adaptability and so on. At the current stage, in the research on indoor positioning technology, most concentrated in the localization algorithm, but to establish the radio map is the basis of localization algorithm, it will greatly affects the positioning system’s performance. In this paper, RF fingerprint map has been researched, and the main work and contributions are as follows:(1)For training workload and time-consuming problem of establishing the radio map, this paper has in-depth researched and analyzed the indoor radio frequency signal propagation characteristics and affect factors. Through sampling the signal strength of the portion interior mesh centre, using radial basis function neural networks prediction method, predicting the signal strength of the remaining grid center, we can quickly build out the entire radio map. Then we use the K-nearest neighbor algorithm(k=3) online positioning to test the performance of this method. Experiments show that, compared with the linear interpolation method, when the sampling rate is less than 50%, the former has better positioning performance. That is to say, this method can greatly reduce the training effort without reducing the positioning accuracy.(2)Due to the static radio map is not well adapt to environmental changes and other issues, this paper proposes the dynamic radio map building method marked with environment state parameters based on the beacon nodes. This method uses the existing beacon nodes to monitor the changes of the indoor environment without adding any hardware cost, at the on-line locating stage, environmental conditions parameter value decides the update strategy of the radio map. When the environment state parameter C is less than or equal to 1.33, we use linear dynamic compensation method to compensate for the RSS values then match the location; when the environment state parameter C is more than 1.33, we use reconfigurable update method to update the radio map. Comparative results show that, the average location error of dynamic radio map has reduced by 25.7% at most compared with that of static radio map, and the former performances a better environment adaptability.