Particle Filter And Its Applications In Wireless Communications

Optimal bayesian estimation is commonly used to calculate the posterior probabilitydensity distribution of the target state, in linear gaussian system,the mean and variance ofstate can be obtained by recursive kalman filtering, however, for non linear dynamic systems,we can’t get the accurate analytical solutions of posterior distribution due to the presence ofmultiple integral, so there are many approximate suboptimal algorithm is proposed to solvethis problem, in these optimization algorithms, particle filter has the best estimation precisionand has the best development prospects.As the main methods of state estimation in the nonlinear dynamic system, particle filteris a kind of through the nonparametric of the monte carlo simulation method to solve therecursive bayesian filter,it is using a set of particles of the state space to update according tobayes rule to estimate the posteriori probability density of the unknown states, when thesample particles tends to infinity, then the probability density which is approximated can beequivalent to the true posterior probability density. For the State estimation of nonlinearsystem, particularly nonlinear non-gaussian systems, particle filter has provided a simple andeffective tools,it has received more and more attention for its simple design, programmingand easy implementation, it has been widely used in communications, radar, navigation, targettracking and machine learning, and other fields.Although particle filter has got many gratifying progress both at home and abroad, it’sdevelopment time is not long and has many key technical problems are still unresolved, suchas how to build the proposal distribution function,the sample depletion which is caused byresampling algorithm, weights degradation phenomenon, the algorithm convergence problems.This article start with the study of the optimal Bayesian filter method,from linear gaussfiltering method gradual transition to the particle filter algorithm,mainly studies the proposedistribution function and resampling technique,and then two kinds of improved particle filteralgorithm is proposed. One of them is from the perspective of differential evolution algorithm,by introducing the adaptive strategy,managing these sample with a way of adaptive variationand adaptive differential hybrid to increase the diversity of particles,and then solved theproblem of the samples exhausted which is caused by resampling;another improved algorithmstart from the particle swarm optimization and artificial immune algorithm, combining theadvantages of both methods to optimizing process these samples,expanding the scope of theoptimization of the algorithm and make the particles move to the high likelihood area,anesising the the weights of degradation which is caused by the difference between proposaldistribution function and the true posterior distribution. Simulation experiments show thatthese two kinds of improved particle filter algorithm can effectively alleviate the particle weight degradation, shorten the running time of the algorithm, improving the precision ofstate estimation. Finally, to improve the performance of signal transmission in wirelesscommunication, one of the improved particle filter algorithm is used to estimate the channelof wireless communication systems.