Research On RF Indoor Location Algorithm Based On LANDMARC And Compressive Sensing

Indoor location is very close to us. In recent years, our casualty and property loss has been reduced in a large extent because that the Radio Frequency Identification indoor location technology is widely used in the mine, hospitals, prisons and other places. However, there are always some contradictions between the localization accuracy and the operational efficiency in the existing RFID indoor location algorithm, which is a technical bottleneck for us to be further studied.In this paper, LANDMARC(Location Identification based on Dynamic Active RFID Calibration) and Compressive Sensing was used to indoor location after analyzing a lot of literature and summarizing the advantages and disadvantages of various indoor location algorithm. And Simulated Annealing algorithm was used to improve present algorithm. And we have done the following work:1. The key of indoor location technology based on the Received Signal Strength Indication is the establishment of wireless signal indoor propagation loss model. The location accuracy will be directly affected by the model. Because of the indoor environment is complex, the traditional method based on parameter fitting to establish the propagation loss model is easily affected by the multipath effect. And it is always bringing poor environment adaptability and large location error. Moreover, many problems has been found about the use of Artificial Neural Network and Support Vector Machine learning algorithm in propagation loss model training, such as, too much number of samples, long training time, the parameters must be adjusted, the simple model is not easy to obtain, and so on. In view of this, Relevance Vector Machine was put forward for loss model training, and to further enhance the learning ability of RVM, a multi kernel function was designed as well. In addition, a data preprocessing and filtering algorithm based on extended Simulated Annealing was put forward to preprocess samples before training. Since the algorithm can keep balance of noise reducing and useful signal reserving, it can reduce the impact of sample error. It was showed by the simulation and experimental results that the proposed algorithm can obtain a good adaptability to the environment, and can also provide technical support for realization of indoor location.2. A double stage location algorithm based on LANDMARC and Compressed Sensing was put forward, in consideration of the contradiction between the location accuracy and computational efficiency of the previous indoor location algorithm. First of all, LANDMARC location algorithm was used in regional location, to lock the target area quickly; then in the locked area, Compressed Sensing theory was introduced. In this part, the scale of Compressed Sensing measurement matrix was determined by constructing virtual reference tags according to the scale of the locked area. And then, the measurement matrix was constructed by using signal strength data of virtual reference tags, which came from a propagation loss model. By the way, Relevance Vector Machine of mixed kernel functions was used to get this model here. At last, Regularized Orthogonal Matching Pursuit reconstruction algorithm was used to get the position index matrix, and the position information of the target can also be obtained through a simple weighted average calculation. In order to improve the location accuracy, Simulated Annealing algorithm was added to modify the iterative factor to match the sparsity. It was showed by the experimental results that the average location error of the proposed algorithm is efficient and good enough to meet the requirements of indoor location.3. A wireless indoor location system has been built by using of RF device with CC2530 as the core. The algorithm was verified by continuous experiments and improvement, so that it could provide a reference for the practical applications.