The design, analysis and parallel implementation of particle filter(PF) were investigated. Firstly, to tackle the particle degeneracy problem in the PF, an iterated importance density function(IIDF) was proposed, wher...The design, analysis and parallel implementation of particle filter(PF) were investigated. Firstly, to tackle the particle degeneracy problem in the PF, an iterated importance density function(IIDF) was proposed, where a new term associating with the current measurement information(CMI) was introduced into the expression of the sampled particles. Through the repeated use of the least squares estimate, the CMI can be integrated into the sampling stage in an iterative manner, conducing to the greatly improved sampling quality. By running the IIDF, an iterated PF(IPF) can be obtained. Subsequently, a parallel resampling(PR) was proposed for the purpose of parallel implementation of IPF, whose main idea was the same as systematic resampling(SR) but performed differently. The PR directly used the integral part of the product of the particle weight and particle number as the number of times that a particle was replicated, and it simultaneously eliminated the particles with the smallest weights, which are the two key differences from the SR. The detailed implementation procedures on the graphics processing unit of IPF based on the PR were presented at last. The performance of the IPF, PR and their parallel implementations are illustrated via one-dimensional numerical simulation and practical application of passive radar target tracking.展开更多
Particle filters have been widely used in nonlinear/non- Gaussian Bayesian state estimation problems. However, efficient distribution of the limited number of particles （n state space remains a critical issue in desi...Particle filters have been widely used in nonlinear/non- Gaussian Bayesian state estimation problems. However, efficient distribution of the limited number of particles （n state space remains a critical issue in designing a particle filter. A simplified unscented particle filter （SUPF） is presented, where particles are drawn partly from the transition prior density （TPD） and partly from the Gaussian approximate posterior density （GAPD） obtained by a unscented Kalman filter. The ratio of the number of particles drawn from TPD to the number of particles drawn from GAPD is adaptively determined by the maximum likelihood ratio （MLR）. The MLR is defined to measure how well the particles, drawn from the TPD, match the likelihood model. It is shown that the particle set generated by this sampling strategy is more close to the significant region in state space and tends to yield more accurate results. Simulation results demonstrate that the versatility and es- timation accuracy of SUPF exceed that of standard particle filter, extended Kalman particle filter and unscented particle filter.展开更多
In order to resolve the state estimation problem of nonlinear/non-Gaussian systems, a new kind of quadrature Kalman particle filter （QKPF） is proposed. In this new algorithm, quadrature Kalman filter （QKF） is used...In order to resolve the state estimation problem of nonlinear/non-Gaussian systems, a new kind of quadrature Kalman particle filter （QKPF） is proposed. In this new algorithm, quadrature Kalman filter （QKF） is used for generating the impor- tance density function. It linearizes the nonlinear functions using statistical linear regression method through a set of Gaussian- Hermite quadrature points. It need not compute the Jacobian matrix and is easy to be implemented. Moreover, the importantce density function integrates the latest measurements into system state transition density, so the approximation to the system poste- rior density is improved. The theoretical analysis and experimen- tal results show that, compared with the unscented partcle filter （UPF）, the estimation accuracy of the new particle filter is improved almost by 18%, and its calculation cost is decreased a little. So, QKPF is an effective nonlinear filtering algorithm.展开更多
基金Project(61372136) supported by the National Natural Science Foundation of China
文摘The design, analysis and parallel implementation of particle filter(PF) were investigated. Firstly, to tackle the particle degeneracy problem in the PF, an iterated importance density function(IIDF) was proposed, where a new term associating with the current measurement information(CMI) was introduced into the expression of the sampled particles. Through the repeated use of the least squares estimate, the CMI can be integrated into the sampling stage in an iterative manner, conducing to the greatly improved sampling quality. By running the IIDF, an iterated PF(IPF) can be obtained. Subsequently, a parallel resampling(PR) was proposed for the purpose of parallel implementation of IPF, whose main idea was the same as systematic resampling(SR) but performed differently. The PR directly used the integral part of the product of the particle weight and particle number as the number of times that a particle was replicated, and it simultaneously eliminated the particles with the smallest weights, which are the two key differences from the SR. The detailed implementation procedures on the graphics processing unit of IPF based on the PR were presented at last. The performance of the IPF, PR and their parallel implementations are illustrated via one-dimensional numerical simulation and practical application of passive radar target tracking.
基金supported by the National Natural Science Foundation of China(61271296)
文摘Particle filters have been widely used in nonlinear/non- Gaussian Bayesian state estimation problems. However, efficient distribution of the limited number of particles （n state space remains a critical issue in designing a particle filter. A simplified unscented particle filter （SUPF） is presented, where particles are drawn partly from the transition prior density （TPD） and partly from the Gaussian approximate posterior density （GAPD） obtained by a unscented Kalman filter. The ratio of the number of particles drawn from TPD to the number of particles drawn from GAPD is adaptively determined by the maximum likelihood ratio （MLR）. The MLR is defined to measure how well the particles, drawn from the TPD, match the likelihood model. It is shown that the particle set generated by this sampling strategy is more close to the significant region in state space and tends to yield more accurate results. Simulation results demonstrate that the versatility and es- timation accuracy of SUPF exceed that of standard particle filter, extended Kalman particle filter and unscented particle filter.
基金supported by the National Natural Science Foundation of China(60574033)
文摘In order to resolve the state estimation problem of nonlinear/non-Gaussian systems, a new kind of quadrature Kalman particle filter （QKPF） is proposed. In this new algorithm, quadrature Kalman filter （QKF） is used for generating the impor- tance density function. It linearizes the nonlinear functions using statistical linear regression method through a set of Gaussian- Hermite quadrature points. It need not compute the Jacobian matrix and is easy to be implemented. Moreover, the importantce density function integrates the latest measurements into system state transition density, so the approximation to the system poste- rior density is improved. The theoretical analysis and experimen- tal results show that, compared with the unscented partcle filter （UPF）, the estimation accuracy of the new particle filter is improved almost by 18%, and its calculation cost is decreased a little. So, QKPF is an effective nonlinear filtering algorithm.
