【摘要】 传统EKF、UKF、粒子滤波算法在解决航空无人机导航定位非线性问题时存在误差大、定位估计精度低等问题,提出了一种核函数正则粒子滤波算法,选取近地面航行,观测x,y方向的位置,并对SLAM非线性模型进行估计.实验数据表明,采用核函数正则粒子滤波算法由于保持了采样粒子的多样性与代表性,保证了在给定模型参数初值下,模型对载体速度和位置信息的跟踪估计能力,其精度比扩展卡尔曼算法的滤波精度高很多;另外,新算法对姿态角估计误差均收敛于0°～1°范围,之后趋近于0°.对于传统滤波算法对载体的航向角误差估计,在整个仿真时间内,其误差值均大于核函数正则算法的误差估计.新算法较传统粒子滤波算法,其滤波精度较高,且算法稳定性与收敛性更强.更多还原

【Abstract】 The traditional EKF, UKF and particle filter algorithms have the problems of large error and low positioning accuracy when solving the nonlinear problem of the UAV navigation positioning. A kernel function regular particle filter algorithm is proposed for the SLAM nonlinear model. The near-surface navigation is selected to observe the position in the x and y directions and the nonlinear model of SLAM was estimated. The experimental data showed that the kernel function regular particle filter algorithm maintains the diversity and representativeness of the sampled particles, and guarantees the ability of the model to track and estimate the carrier speed and position information under the initial value of the given model parameters, and its filtering accuracy much higher than EKF. What’s more, the estimate error of the attitude angle with the new algorithm converges within 0°～1°, and then approaches 0°. For the traditional filter algorithm, the attitude angle error value is larger than the kernel function regular algorithm in the whole simulation time. Compared with the traditional particle filter algorithm, the new algorithm has higher filtering precision and stronger algorithm stability and convergence.更多还原