针对非线性、非高斯系统状态的在线估计问题,提出一种新的基于序贯重要性抽样的粒子滤波算法.在滤波算法中,用状态参数分解和退火系数来产生重要性概率密度函数,此概率密度函数综合考虑了转移先验、似然、噪声的统计特性以及最新的观察数据,因此更接近于系统状态的后验概率.理论分析与仿真实验表明该粒子滤波器的性能明显优于标准的粒子滤波器和扩展卡尔曼滤波器. Aiming at the problem of on-line estimation of nonlinear and non-Gaussian system states, a new particle filter algorithm based on sequential importance sampling is proposed. In the filtering algorithm, state parameter decomposition and annealing coefficients are used to generate the importance probability density function The probability density function considers the statistical properties of the transfer priori, likelihood, noise and the latest observation data, so it is closer to the posterior probability of the system state. Theoretical analysis and simulation show that the performance of the particle filter is better than the standard Particle Filter and Extended Kalman Filter.