随着无人系统、自动驾驶等领域的快速发展，双动态载体相对定位技术在满足高精度实时定位和复杂环境适应性方面变得愈加重要。为提升现有双动态载体相对定位算法的精度与可靠性，构建了参考站处理、模糊度固定、移动站解算优化方法，提出了一种双重因子图与模糊度优化（DF-AR）的双动态载体相对定位算法。利用融合多频多系统卡尔曼滤波的因子图优化模型抑制参考站端单点定位模式的误差，提升相对定位精度。并通过基线约束与数据质量定权，构建了改进的数据加模型驱动部分模糊度固定策略，优选出可靠性更高的模糊度子集，提升模糊度固定成功率和相对定位解的可靠性。在上述改进基础上，在因子图优化模型中引入滑动窗口，动态调整数据量，对移动站定位解进行再优化，获得了更为鲁棒的相对定位结果。分别开展了静态评估实验、双车载以及无人机/车载动态相对定位实验。实验结果表明：在不同的实验场景下，DF-AR算法相比于RTKLIB算法的基线误差分别降低了69.72%、94.89%和68.03%，基线解精度由米级提升至分米级，有效提高了相对定位的可靠性和精确性。

With the rapid development of unmanned systems， autonomous driving， and other related fields， dual dynamic carrier relative positioning technology has become increasingly important for achieving high-precision real-time positioning and adapting to complex environments. To improve the accuracy and reliability of the existing relative positioning algorithm for dual dynamic carrier， this paper proposes a dual dynamic carrier relative positioning algorithm based on Double Factor graph and Ambiguity Resolution optimization （DF-AR）， incorporating reference station processing， ambiguity resolution， and mobile station resolution optimization methods. To improve relative positioning accuracy， a factor graph optimization model integrating multi-frequency and multi-system Kalman filtering is used to suppress single-point positioning errors at the reference station. Using baseline constraints along with data quality weighting， an improved data and model-driven partial ambiguity resolution strategy is constructed. The ambiguity subset with higher reliability is selected to improve the success rate of ambiguity fixation and the reliability of the relative positioning solution. Based on these improvements， a sliding window is introduced in the factor graph optimization model to dynamically adjust the data amount. The positioning solution of the mobile station is reoptimized to achieve more robust relative positioning results. Static evaluation experiments， dual-vehicle and UAV/vehicle dynamic relative positioning experiments were carried out. The experimental results show that in different experimental scenarios， the baseline solution error of the DF-AR relative positioning algorithm has an error reduction of 69.72%， 94.89%， and 68.03% compared to the RTKLIB algorithm. The baseline solution accuracy has been improved from meter level to decimeter level， effectively enhancing the reliability and accuracy of relative positioning.