Study On The Localization Algorithms In Wireless Sensor Networks Based On Hidden Markov Model

In recent years, the emerging technology--wireless sensor network has received extensive attention from the international academia and industry, which has a broad prospect in the fields of military, environment, industry and so on. The basic function of wireless sensor networks is to calculate the location information of mobile node.In both the field of disaster relief and of smart home, how to accurately measure the coordinate information of the mobile nodes is of great importance to the entire positioning system. In this thesis, a strategy is studied to eliminate the None-Line-Of-Sight error (NLOS) in the unfamiliar indoor environment, wireless localization algorithms are presented and the corresponding performance is analyzed.According to the disadvantage of D/TA (Detection/Tracking Algorithm and its motion feature, the localization algorithm is upgraded to Improved-Detection/Tracking Algorithm (I-D/TA) combining with the motion inertia. To improve HMM algorithm, another two modified algorithms M-HMM and RM-HMM are presented. Simulation experiment showed that these three algorithms mentioned above have a good performance in both distance estimation and coordinate calculation. The accuracy of tree algorithms increases one by one. And all the algorithms are stable.According to the speed feature of the mobile node, a hybrid localization algorithm is presented combined with HMM and IMM. The velocity model is divided into high speed model and low speed model. Each moving node assess their probability of the two states continually and use the IMM model to integrate the results from HMM and modified forms of HMM in order to achieve a better precision. Simulation experiment showed that the hybrid localization algorithm combined with HMM and IMM get excellent effects in distance estimation and coordinate calculation. The final location accuracy is higher than Kalman Filter, Particle Filter and other algorithms. The result is stable. The results of the experiment showed that this algorithm has good robustness.On account of the different environment, the initial values of the location algorithm based on HMM are different, a hybrid optimization algorithm based on Particle Swarm Optimization (PSO) and Simulated Annealing (SA) is presented to optimize the initial value of HMM. In addition, in order to speed up the efficiency without reduction of accuracy, an optimization strategy is used with dimensionality reduction. The traditional way of boundary treatment tends to produce a local optimal solution which should be eliminated. So taking that into account, two improved strategies are proposed to process the boundary. The result of the computer simulation experiments shows that, compared with other method, the initial value optimizing strategy has a higher calculating speed and a more stable precision.