一种快速解算高维模糊度的LLL分块处理算法

摘要

由于多频多模GNSS观测数据解算的模糊度具有较高的维数和精度,当采用常规的LLL算法进行模糊度整数估计时,规约耗时显著大于搜索耗时,成为限制高维模糊度解算计算效率的主要因素.针对这一问题,通过分析规约耗时与模糊度维数和精度之间的关系,提出了一种 LLL 分块处理算法.该算法通过对模糊度方差协方差阵进行分块处理,降低单个规约矩阵的维数,以减少规约耗时,从而提高模糊度解算计算效率.通过两组实测高维模糊度数据对本文提出的分块处理算法进行了效果验证.结果显示,当分块选择合理时,本文提出的算法相对于 LLL算法的解算效率分别可提高 65.2%和 60.2%.%Due to high dimension and precision for the ambiguity vector under GNSS observations of multi-frequency and multi-system,a major problem to limit computational efficiency of ambiguity resolution is the longer reduction time when using conventional LLL algorithm.To address this problem,it is proposed a new block processing algorithm of LLL by analyzing the relationship between the reduction time and the dimensions and precision of ambiguity.The new algorithm reduces the reduction time to improve computational efficiency of ambiguity resolution,which is based on block processing ambiguity variance-covariance matrix that decreased the dimensions of single reduction matrix.It is validated that the new algorithm with two groups of measured data.The results show that the computing efficiency of the new algorithm increased by 65.2%and 60.2% respectively compared with that of LLL algorithm when choosing a reasonable number of blocks.