无味卡尔曼滤波算法形式及性能研究

The Study on Form and Performance of Unscented Kalman Filtering Algorithm

通常认为当系统噪声与量测噪声为加性时,是否将噪声扩展为状态量并不影响无味卡尔曼滤波算法性能。针对这种观点,文中利用变尺度对称集无味变换,在复杂加性噪声模型下,推导并证明了两者的差异,说明了上述观点的不全面性。并通过对扩展与非扩展、重采样与非重采样组合的四种算法形式仿真,研究分析了四种形式下算法性能的差异。结果表明状态扩展有利于提高无味卡尔曼滤波算法性能,从而证明了理论分析的正确性。

It is usually believed that unscented Kalman filter has the same performance whether in extending state or not when system noise and measurement noise are additive. To be against this view, the scaled symmetric set unscented transformation was used to deduce and prove the difference between them. Four kinds of UKF including extension, non-extension, resample and non-resample were simulated and analyzed. The results indicate that there are differences between four kinds of UKF algorithm, the state extension is benefit for improving UKF performance, which proves the correctness of theory analysis.