为了提高非线性变换的近似精度,提出了一种高阶无迹变换(High orderUnscented Transform,HUT)机制,利用HUT确定采样点并进行数值积分去近似状态的后验概率密度函数,建立了高阶无迹卡尔曼滤波(High-order UnscentedKalman Filter,HUKF)算...为了提高非线性变换的近似精度,提出了一种高阶无迹变换(High orderUnscented Transform,HUT)机制,利用HUT确定采样点并进行数值积分去近似状态的后验概率密度函数,建立了高阶无迹卡尔曼滤波(High-order UnscentedKalman Filter,HUKF)算法.进一步的为了解决非线性、非高斯系统的状态估计问题,将HUKF与高斯和滤波(Gaussian Sum Filter,GSF)相结合,提出了一种高斯和高阶无迹卡尔曼滤波算法(Gaussian Sum High order Unscented Kalman filter,GS-HUKF),该算法的核心思想是利用一组高斯分布的和去近似状态的后验概率密度,同时针对每一个高斯分布采用高阶无迹卡尔曼滤波算法进行估计.数值仿真实验结果表明,提出的HUT机制与普通的无迹变换(Unscented Transform,UT)相比,具有更高的近似精度;提出的GS-HUKF与传统的GSF以及高斯和粒子滤波器(Gaussian Sum Particle Filter,GS-PF)相比,兼容了二者的优点,即具有计算复杂度低和估计精度高的特性.展开更多
For being able to deal with the nonlinear or non-Gaussian problems, particle filters have been studied by many researchers. Based on particle filter, the extended Kalman filter （EKF） proposal function is applied to ...For being able to deal with the nonlinear or non-Gaussian problems, particle filters have been studied by many researchers. Based on particle filter, the extended Kalman filter （EKF） proposal function is applied to Bayesian target tracking. Markov chain Monte Carlo （MCMC） method, the resampling step, ere novel techniques are also introduced into Bayesian target tracking. And the simulation results confirm the improved particle filter with these techniques outperforms the basic one.展开更多
Combining interacting multiple model （IMM） and unscented particle filter （UPF）, a new multiple model filtering algorithm is presented. Multiple models can be adapted to targets＇ high maneu- vering. Particle filte...Combining interacting multiple model （IMM） and unscented particle filter （UPF）, a new multiple model filtering algorithm is presented. Multiple models can be adapted to targets＇ high maneu- vering. Particle filter can be used to deal with the nonlinear or non-Gaussian problems and the unscented Kalman filter （UKF） can improve the approximate accuracy. Compared with other interacting multiple model algorithms in the simulations, the results demonstrate the validity of the new filtering method.展开更多
A new particle filter is presented for nonlinear tracking problems. Inpractice, maneuvering target-tracking systems are usually nonlinear and incompletely observed, andthe main difficulty of maneuvering target-trackin...A new particle filter is presented for nonlinear tracking problems. Inpractice, maneuvering target-tracking systems are usually nonlinear and incompletely observed, andthe main difficulty of maneuvering target-tracking problem lies in the fact that the maneuverabilityat every step is of high uncertainties. Here a new smoothing particle filter algorithm is proposed,which combines the particle filter to tackle the non-linear and non-Gaussian peculiarities of theproblem, together with smoothing of the PDF of system modes and thus settles the estimate problem ofthe target maneuverability. The simulation comparison with the auxiliary particle filters showsthat the approach has superiority and yields performance improvements in solving nonlinear trackingproblems.展开更多
针对组合导航系统为高维非线性非高斯的特点,提出了一种在线实时调整粒子个数的自适应MCMC(Markov Chain Monte Carlo)粒子滤波算法。该算法利用基于KLD(Kullback-Leibler Distance)采样方法的自适应粒子个数调整算法在线调整MCMC粒子...针对组合导航系统为高维非线性非高斯的特点,提出了一种在线实时调整粒子个数的自适应MCMC(Markov Chain Monte Carlo)粒子滤波算法。该算法利用基于KLD(Kullback-Leibler Distance)采样方法的自适应粒子个数调整算法在线调整MCMC粒子滤波过程中的粒子个数,利用预测粒子在状态空间中的分布情况来在线实时的确定下一次滤波迭代所需的粒子个数,从而有效减小算法的运算量,提高MCMC粒子算法的实时处理能力。最后,将该算法应用到了组合导航系统中进行了仿真研究。通过仿真结果可以看出,该算法在保持了MCMC粒子滤波算法的估计能力的同时,有效降低了算法的计算量,更适合于实际应用。展开更多
文摘为了提高非线性变换的近似精度,提出了一种高阶无迹变换(High orderUnscented Transform,HUT)机制,利用HUT确定采样点并进行数值积分去近似状态的后验概率密度函数,建立了高阶无迹卡尔曼滤波(High-order UnscentedKalman Filter,HUKF)算法.进一步的为了解决非线性、非高斯系统的状态估计问题,将HUKF与高斯和滤波(Gaussian Sum Filter,GSF)相结合,提出了一种高斯和高阶无迹卡尔曼滤波算法(Gaussian Sum High order Unscented Kalman filter,GS-HUKF),该算法的核心思想是利用一组高斯分布的和去近似状态的后验概率密度,同时针对每一个高斯分布采用高阶无迹卡尔曼滤波算法进行估计.数值仿真实验结果表明,提出的HUT机制与普通的无迹变换(Unscented Transform,UT)相比,具有更高的近似精度;提出的GS-HUKF与传统的GSF以及高斯和粒子滤波器(Gaussian Sum Particle Filter,GS-PF)相比,兼容了二者的优点,即具有计算复杂度低和估计精度高的特性.
基金This project was supported by the National Natural Science Foundation of China (50405017) .
文摘For being able to deal with the nonlinear or non-Gaussian problems, particle filters have been studied by many researchers. Based on particle filter, the extended Kalman filter （EKF） proposal function is applied to Bayesian target tracking. Markov chain Monte Carlo （MCMC） method, the resampling step, ere novel techniques are also introduced into Bayesian target tracking. And the simulation results confirm the improved particle filter with these techniques outperforms the basic one.
文摘Combining interacting multiple model （IMM） and unscented particle filter （UPF）, a new multiple model filtering algorithm is presented. Multiple models can be adapted to targets＇ high maneu- vering. Particle filter can be used to deal with the nonlinear or non-Gaussian problems and the unscented Kalman filter （UKF） can improve the approximate accuracy. Compared with other interacting multiple model algorithms in the simulations, the results demonstrate the validity of the new filtering method.
文摘A new particle filter is presented for nonlinear tracking problems. Inpractice, maneuvering target-tracking systems are usually nonlinear and incompletely observed, andthe main difficulty of maneuvering target-tracking problem lies in the fact that the maneuverabilityat every step is of high uncertainties. Here a new smoothing particle filter algorithm is proposed,which combines the particle filter to tackle the non-linear and non-Gaussian peculiarities of theproblem, together with smoothing of the PDF of system modes and thus settles the estimate problem ofthe target maneuverability. The simulation comparison with the auxiliary particle filters showsthat the approach has superiority and yields performance improvements in solving nonlinear trackingproblems.
文摘针对组合导航系统为高维非线性非高斯的特点,提出了一种在线实时调整粒子个数的自适应MCMC(Markov Chain Monte Carlo)粒子滤波算法。该算法利用基于KLD(Kullback-Leibler Distance)采样方法的自适应粒子个数调整算法在线调整MCMC粒子滤波过程中的粒子个数,利用预测粒子在状态空间中的分布情况来在线实时的确定下一次滤波迭代所需的粒子个数,从而有效减小算法的运算量,提高MCMC粒子算法的实时处理能力。最后,将该算法应用到了组合导航系统中进行了仿真研究。通过仿真结果可以看出,该算法在保持了MCMC粒子滤波算法的估计能力的同时,有效降低了算法的计算量,更适合于实际应用。
