Abstract: For the sake of effect fusion of multi-sensor redundant system metrical information，making the value of state estimation approach to the true value and retaliating high accuracy and high reliability state estimation，three algorithms，based on optimally weighted least square method（OW-LSM）and finite windowing weighted algorithm（FW-LSM）and the self-learning weighted least squares（SL-LSM），are applied to information fusion of multi-sensor data respectively.The mean variance of data is reduced，estimation accuracy is advanced dramatically.Then the simulation comparison with traditional least squares is performed.Results show that the three algorithms has higher accuracy than the traditional one.Therein，Self-learning weighted least squares taken the effect of history data and ambient noise and new sample value into account，enhanced sensitivity to noise measure，has better estimation effect.

Key words: least square method, multi-sensor data fusion, data fusion

摘要： 为了有效融合多传感器冗余系统量测信息，使状态的估计值更接近于状态的真实值，实现高精度和高可靠性的状态估计，采取了基于最优加权的最小二乘算法、有限窗加权的最小二乘算法和自学习加权最小二乘算法，分别对多传感器实测数据进行融合处理，融合后数据的方差大幅度降低，估计精度显著提高。并与传统的最小二乘算法进行了仿真对比，结果表明，这3种方法较最小二乘算法融合精度更高，其中，自学习加权的最小二乘融合算法既考虑了历史数据的作用，又考虑了环境噪声和新的采样值的影响，增强了对噪声检测的敏感性，估计效果较好。

关键词: 最小二乘, 多传感器数据融合, 数据融合