基于广义最小二乘模型,建立了一种带滑动窗的动态OD矩阵估计算法,可通过对路段交通量和行程时间的检测来估计时变的OD数据.对模型中关键的交通分配矩阵,给出了解析的计算公式.算法是一种递推的估计过程,仅需较少的先验信息,且估计过程不会发散;滑动窗的引入可充分利用量测信息,抑制量测噪声. Based on the generalized least squares model, a dynamic OD matrix estimation algorithm with sliding window is proposed to estimate the time-varying OD data by detecting the traffic volume and travel time of the road segment. For the key traffic assignment matrix in the model, The algorithm is a recursive estimation process, which requires less prior information and the estimation process will not diverge. The introduction of sliding window can make full use of the measurement information to suppress the measurement noise.