中文摘要：

交通需求组合模型（CTDM）基于随机效用理论组合了出行-分布-方式-路径选择，突破了传统四步骤模型的局限，并可以描述为一个非线性规划问题．对求解CTDM的部分线性化算法进行分析，提出了在进行步长优化时使用二次插值法得到模拟最优步长的方法．使用经典路网Sioux Falls，将所提出的算法与连续平均法和精确线性搜索算法比较，验证了二次插值法的计算精度高，收敛速度快．

英文摘要：

The combined travel demand model （CTDM） combines the travel-destination-mode-route choice based on the random utility theory. This model avoids the limitation of the conventional sequential four-step procedure, and can be formulated as a non-linear programming problem. By analyzing the partial linearization algorithm of CTDM, a quadratic interpolation method is proposed to obtain the approximated optimal step size. Comparison of the proposed algorithm with method of successive averages and bi-section line search method in the classic Sioux Falls network confirms that the quadratic interpolation can converge faster and get better solution than the other two methods.