【作者】 聂莉娟；

【导师】 吴俊伟；

【作者基本信息】 哈尔滨工程大学 ， 导航、制导与控制， 2004， 硕士

【摘要】 初始对准的精度直接关系到惯导系统的工作精度，初始对准的时间是惯导系统的重要战术技术指标。因此，初始对准是惯导系统最重要的关键技术之一。 目前，卡尔曼滤波是捷联惯导系统初始对准的最成熟、应用最广泛的滤波方法。但是，由于卡尔曼滤波要求干扰是统计特性已知的白噪声，该条件在实际应用中通常不能严格满足，由此影响到滤波的性能。H_∞滤波要求已知噪声是能量有界的，而不是其精确的统计特性。H_∞滤波器不但对噪声的不确定性具有鲁棒性，而且对系统参数的不确定性也具有鲁棒性的能力，用H_∞滤波能够在上述不确定的情况下完成滤波计算，因此本文在系统噪声或量测噪声为有色噪声或存在模型失配时，就用此滤波器代替卡尔曼滤波器进行初始对准。 本文用MATLAB语言分别实现了捷联惯导初始对准的卡尔曼滤波器以及H_∞滤波器的设计，并将设计结果进行仿真。仿真结果表明：在理想条件下，当噪声统计特性已知的情况下，卡尔曼滤波的精度将优于H_∞滤波；而当建模误差存在或者是系统模型统计特性未知时，卡尔曼滤波器不能完成滤波计算，而H_∞滤波仍能保持较高的精度。采用H_∞滤波器进行随机状态估计，其特点是：不仅速度快、鲁棒性好、精度较高，而且更符合应用的实际情况。更多还原

【Abstract】 The original alignment precision impacts directly on the precision of the INS (inertial navigation system). The time used in original alignment processing is an important military tactics technique index of the INS. So the original alignment is one of the most important key technical of the INS.At present, the Kalman filter is the most well rounded and widely used filter. Kalman filter requires the known statistical character of system disturbance, which can’t be strictly satisfied in practice system usually. H∞ filtering requires the disturbance is energy bounded other than known accurate statistic. H∞ filter has robustness both in uncertain noise and in uncertain system parameter, and it can be completed in the conditions mentioned above. In this paper, Kalman filter is replaced by H∞ filter in the case of the system model existing error, such as the system being with non-white noise and uncertain parameters.In this paper, the MATLAB is used to realize a Kalman filter and aH∞ filter of the initial alignment of strap-down inertial navigation systems. We contrast the two simulation result. In ideal condition, Kalman filter can keep less error precision in the initial alignment than H∞ filter. When the system model exist error, Kalman filter can not complete filtering, but H∞ filter can keep upperprecision all the me .The advantages of H∞ filter are more rapid, more robust, more precision and it can be more applicable in engineering.更多还原