非平稳非高斯测量噪声条件下改进差分粒子滤波算法研究

针对非平稳非高斯测量噪声(NSNGN)条件下差分粒子滤波(DDPF)算法状态估计精度低、易发散的问题，提出了一种改进DDPF(IDDPF)算法. IDDPF算法采用高斯混合密度函数近似估计测量噪声，替代传统算法中测量噪声的高斯密度函数近似估计，采用似然函数的对数最大化法求解高斯混合密度函数模型参数，并将该模型应用于粒子权值计算，避免了高斯密度函数近似估计噪声模型所易于导致的粒子退化问题；通过建立水下目标纯方位角跟踪系统模型，将IDDPF算法应用于闪烁测量噪声条件下水下目标纯方位角跟踪问题的求解。50次Monte Carlo对比仿真实验结果表明：在NSNGN条件下IDDPF算法具有跟踪响应快、估计精度高、鲁棒性较好等优点。

Research on Improved Divided Difference Particle Filter under Non-stationary Non-Gaussian Noise Background

An improved divided difference particle filter (IDDPF) is proposed to improve the low accuracy of state estimation and divergent tend problem which may be caused by divided difference particle filter (DDPF) in the condition of the non-stationary non-Gaussian measurement noise (NSNGN). The IDDPF algorithm adopts Gaussian mixture density function to approximately estimate the measurement noise, instead of the Gaussian density function usually adopted in DDPF. The noise parameters are estimated by maximizing the log likelihood function of the measurement noise model. The model is then used to calculate the particle weight, which avoids the particle degeneracy problem. The IDDPF algorithm is tested by establishing bearing-only tracking of underwater target under the glint measurement noise background. The results of 50 Monte-Carlo simulation experiments show that the IDDPF algorithm has the advantages of fast tracking response, high estimated precision and robustness, etc. under NSNGN background.