| Wi-Fi based indoor localization has attracted many research efforts owing to the flourish of Wi-Fi device and pervasive computing. As a mainstream measurement, Received Signal Strength Indicator(RSSI) has been adopted in vast indoor localization systems. In recent years, some devices(e.g., Intel 5300 NIC) support extract the Channel State Information(CSI), which is a PHY layer power feature. CSI holds the potential for the convergence of accurate and pervasive indoor localization, thus appealing many researchers’ attention. In this paper, we conduct the indoor localization research based on CSI. The major contributions are outlined as follows:1. An innovative localization scheme called Ran Li Fi(Range-based Lightweight Fingerprint) is proposed, which contains pre-process stage, offline stage and online stage. The pre-process stage is the prerequisite of the ongoing offline and online stage, which handles the raw CSI data and outputs the stable ones. The offline stage is to build the fingerprint map, while the online stage completes the distance measurement and fingerprint match, and finally give the position of the target. By leveraging an innovative annular sampling method, the range-based scheme and the fingerprint based one can be well combined and complement each other. That is, after getting the distance d between the target and the AP, the set of the fingerprints to be matched could be reduced onto the cyclic annular with radius closest to d. Finally, the position corresponding to the fingerprint which has the greatest matching degree is regarded as the position of the target.2. Based on the Bhattacharyya coefficient, a kind of fingerprint named B-Ao E(Bhattacharyya based Average of Energy of Interest) is proposed. By comparing B-Ao E and another 5 fingerprint candidates, we find that B-Ao E is the only one which well satisfied the internal similarity and external distinguishability, i.e., the B-Ao E fingerprints from the same position are extremely similar, while the ones from different positions are differentiable.3. To accurately measure the distance, a Self-adaptive and Energy of Interest basedpropagation model, which is named S-Eo I, is presented. Based on a simple analysis, we can know that the LOS and NLOS signal component cannot be separated with 20 MHz bandwidth. What is more, even in the same environment, the fading level of each signal transmission is different. In view of this, we define energy of interest and a kind of self-adaptive path-loss factor. The energy of interest is stable and has the ability to sense the environment changes, and the path-loss factor can reflect the fading level of each signal transmission. Based on these improvement, we can get a more accurate result.4. An indoor localization system called Ran Li Fi is implemented. The system contains 4 models, i.e., the data acquisition model, pre-process model, fingerprint map model and localization model. The data acquisition model gets the raw CSI data, and the other three models implement the methods in the pre-process stage, offline stage and online stage, respectively. In addition, two advanced mechanism, i.e., fault-tolerant and feedback mechanism, are provided to improve the robustness of the system.5. The performance of Ran Li Fi system is evaluated from many aspects. The mean error of the S-Eo I model is less then 1.7m, and the B-Ao E fingerprint can give the correct postion in 98% cases. With the fault-tolerant mechanism and 1.5m sampling interval, the final localization mean error of Ran Li Fi system is 0.8m. |
PDF Full Text Request