Design And Implementation Of Particle Filter Based On Tabu Immune And Weight Selected Algorithms

As system scale becomes larger, its complexity grows. The original particle filtering algorithms are insufficient in parameter estimation, state estimation as well as target tracking of the system. Therefore, it is markedly important to design a new particle filtering algorithm, to improve the estimation accuracy and to reduce the computational complexity.The particle filtering algorithm was optimized in this paper and the main work was as follows:1. The basic principles, calculation process and some defects of the standard particle filtering algorithm were introduced. Some intelligent algorithms were used such as tabu search algorithm, artificial immune and weight selected algorithm which were connected with the new-designed algorithm.2. On the base of the particle filtering algorithm, a new particle filtering algorithm based on the tabu search and artificial immune algorithm was designed to solve the problem that was particle degradation and low diversity of sample sets. The optimization capability of artificial immune algorithm was used to select the appropriate particle from all particles efficiently, so the diversity of sample sets was greatly improved, and it could avoid falling into local optimum through tabu search. Compared with artificial immune particle filtering and standard particle filtering algorithm, the estimation performance could be verified.3. A marginalized particle filtering algorithm based on weight selected was proposed to solve the problem of high complexity of particle algorithm. The linear substructure of the model was used to reduce the estimation variance from the standard particle filter algorithm. The corresponding linear state variables could be marginalized by this algorithm, and the optimal linear filter was applied to reduce computational complexity significantly. Furthermore, the independence among the particles made the particle set contain more distinct particle paths, so the diversity of particle set was enhanced with better optimization effect.4. The two improved algorithms were used in joint estimation of train braking rate and the train running status under the background of urban rail brake model,and their simulating results were also compared.