基于自适应极大后验估计的空间目标运动状态确定

Adaptive Maximum a Posteriori Estimation for Space Object Dynamic State Determination

提出一种新的基于自适应极大后验(AMAP)估计的空间目标运动状态确定方法,致力于削弱未知干扰对状态估计的不利影响.针对带有干扰的离散时间非线性随机系统设计了AMAP估计算法,采用高斯-牛顿优化方法实现极大后验(MAP)估计,通过模式切换和加权融合强化算法的自适应能力.基于理论分析导出了状态估计均方误差(MSE)的表达式,说明所提算法能够达到优于传统扩展卡尔曼滤波(EKF)和MAP估计算法的精度.以空间目标运动状态确定系统为例,通过蒙特卡洛仿真验证了AMAP估计算法的性能优势,不同条件下的对比研究表明,所提算法具备应对未知干扰的自适应能力,能够有效提升空间目标运动状态估计精度.

This paper presents a novel adaptive maximum a posteriori （AMAP） estimation method for space object dynamic state determination taking into account the influence of model uncertainties. Considering a nonlinear stochastic discrete- time system model with unknown disturbances, the AMAP estimation algorithm adopts the Gauss-Newton iterative optimization steps to implement an approximate maximum a posteriori （MAP） estimation, and the switch-mode combination technique is used to achieve the adaptive capability. The meaJa-square estimation error （MSE） of the state estimate is derived. It is proved that the presented algorithm can yield a smaller MSE than that of the traditional extended Kalman filter （EKF） or iterated extended Kalman filter （IEKF）. The performance advantage of the AMAP estimation algorithm is illustrated via Monte Carlo simulations on a space object dynamic state determination application. Though comparisons in different scenarios, the presented algorithm is shown to improve the adaptability of the filter and ensure the state estimation accuracy.