Tactical Missile Technology

在地形辅助导航中,粒子滤波由于其良好的实时性及对地形免线性化处理的特性而广泛应用。但该滤波算法也面临着粒子退化、高维空间"维数灾难"的问题,为提升导航性能,提出一种融合神经网络的RBPF算法。在SITAN系统基础上,以粒子滤波代替卡尔曼滤波,建立该系统状态空间模型。在粒子滤波重采样步骤前引入反向传播(BP)神经网络,调整奇异点,平滑权值,并采用Rao-Blackwellized (RB)理论对复杂高维模型进行结构分解,保障导航实时性。通过仿真生成飞行轨迹,与真实的地形高程图进行匹配定位,设定不同粒子数目以改变计算量大小,区别仿真时间。仿真实验证明该算法具备良好实时性,能有效改善粒子滤波缺陷,提升导航定位性能。

Particle filter is widely used in terrain aided navigation because of its good real-time performance and non-linearization of terrain. However, this filtering algorithm also faces the problems of particle degradation and“dimensional disaster”in high dimensional space. In order to improve the navigation accuracy, a RBPF algorithm based on neural network fusion is proposed.Based on SITAN system, particle filter is used to replace Kalman filter, and the state space model of SITAN system is established. The Back Propagation(BP) neural network is introduced before the resampling step of particle filter to adjust singular points and smooth weights, and the Rao-Blackwellized(RB) theory is used to decompose the complex high-dimensional model to ensure the real-time navigation. The flight trajectory is generated through simulation,which is matched and positioned with the real terrain elevation map. Different particle numbers are set to change the amount of calculation, and the simulation time is distinguished.The simulation results show that the algorithm has good real-time performance, can effectively improve the defect of particle filtering, and improve the performance of navigation and positioning.