论文部分内容阅读
针对基于扩展卡尔曼滤波(EKF)的航位推算(DR)在系统非线性、噪声非高斯情况下导航精度严重下降的问题,提出一种基于权值方差缩减粒子滤波的航位推算.建立无人水下航行器(UUV)的非线性运动学模型以及传感器的测量模型,利用模拟退火算法的退温函数产生自适应指数渐消因子以降低粒子权值的方差,进而增加有效粒子数,并以此替代标准粒子滤波中的重采样步骤.海试数据仿真试验表明,与基于EKF的航位推算算法相比,所设计算法避免了模型线性化、噪声非高斯的影响;与基于标准粒子滤波的航位推算相比,所设计算法降低了由于重采样导致的粒子贫化程度,从而提高了UUV导航系统的稳定性和精确性.
Aiming at the problem that dead reckoning (DR) based on Extended Kalman Filter (EKF) seriously degrades the navigation accuracy under the condition of nonlinear system and non-Gaussian noise, a dead reckoning based on weight variance reduction particle filter is proposed. Human UUV (UUV) nonlinear kinematics model and sensor measurement model, the annealing function of simulated annealing algorithm is used to generate adaptive exponential fading factor to reduce the variance of particle weights, and then increase the number of effective particles So as to replace the resampling step in standard particle filter.The simulation test of sea trial data shows that compared with the dead reckoning algorithm based on EKF, the designed algorithm avoids the influence of model linearization and noise non-Gaussian, Compared with the dead reckoning, the designed algorithm reduces the degree of particle depletion due to resampling, thus improving the stability and accuracy of the UUV navigation system.