随着经济发展和人民生活水平的提高，汽车销量快速增长，车型品种也变得更加丰富，这对汽车制造提出了更高的挑战。其中，混流生产情形下的车辆排序问题由于问题复杂而备受关注。车辆排序问题主要是为确定工作车间的生产顺序，由于不同车间对生产顺序的偏好不同，企业常在两个连续生产车间之间设置缓冲区实现重排序。设置缓冲区提高了生产调度的灵活度，但问题的求解变得更加的困难，构建高效算法成为研究热点。针对汽车重排序问题，本文构建了带缓冲区的多车间汽车混流生产线联合调度模型，考虑多车间车辆排序决策以及缓冲区的重调度决策，目标是最小化总切换成本。为加速模型求解，本文设计了线性化方法将数学模型转化为混合整数规划模型。在算法设计上，本文设计两阶段迭代优化算法。1）第一阶段是多车间车辆排序，按照车辆类进行建模，同时根据颜色属性推导出问题下界加速求解；之后，通过动态规划算法确认每个颜色类中车辆的生产顺序，获得上游在考虑总装车间目标后的最优解。2）第二阶段是车辆重排序，构建了动态规划算法进行求解，用集合描述状态，弱化模型中栈道对称性的影响，其次为求解大规模问题设计了Rollout算法进行遍历加速，在短时间内找到第二阶段问题的高质量解。为验证本文迭代优化算法的有效性，本文基于文献和真实工业场景生成算例并进行了大量的数值试验。实验结果表明，相较于Gurobi求解器，本文设计的算法在求解效率上表现出显著优势，能在短时间内得到高质量解。有效割对第一阶段问题的加速效果明显，在不影响解质量的情况下平均求解效率提升近200-300倍。第二阶段动态规划问题对模型对称性的消除作用显著，显著缩短求解时间。迭代优化算法中联合求解、内部排序与迭代优化模块均对解质量的提升有较大贡献度，平均对总装车间目标分别提升了90%、38%与45%。

With the rapid development of fuel cell technology, the new energy vehicle industry has gradually become an emerging pillar industry. In the process of development, the rapid growth of automobile production and sales, as well as the wide variety of vehicle models, also pose challenges to traditional automobile manufacturing. Among them, the vehicle scheduling problem in mode-mix production line has attracted much attention due to its significant impact on automotive manufacturing costs and its high complexity. The vehicle sequencing problem is mainly aimed at determining the production sequence of the workshop. Due to the different preferences of different workshops, enterprises often set up buffer zones between two consecutive workshops for requencing, which enables us to further improve the production plans. Therefore, it is of significant urgency to design efficient algorithms to solve the problem.Based on the above background, this paper proposes a joint scheduling problem for multi-workshop mixed flow automobile production lines with buffer zones, jointly considering the vehicle sequencing problem in multiple workshops and the resequencing decision conducted in the buffer zones, and constructing a mathematical model to minimize the total smoothing cost. To accelerate the model solving, this paper designs a linearization method to transform the mathematical model into a mixed integer programming problem. In terms of algorithm design, to solve large-scale instances encountered in the practical application, this paper divides the original problem into two stages and solves it through iterative optimization algorithms to achieve two-stage optimization. 1) The first stage is the multi-workshop vehicle scheduling problem. For the first stage problem, this thesis proposed a vehicle cluster-based model and derives a lower bound to tighen the model by exploiting the color attributes. Afterwards, a dynamic programming algorithm is used to determine the production sequence of vehicles for each vehicle cluster, so that a feasible solution for the upstream problem which also considers the goal of the final assembly workshop is obtained. 2) The second stage is the vehicle reordering problem. For the second stage problem, this thesis innovatively proposed a dynamic programming algorithm, in which sets rather than sequences are used to describe the state, significantly reducing the model‘s symmetry issue. Compared to a commercial solver package, it has a significant acceleration in terms of solving time and solution quality. Secondly, for large-scale problems, this thesis designed a rollout algorithm which adopts the look-ahead philosophy to find high-quality solutions in a short time.To verify the effectiveness of the iterative optimization algorithm proposed in this paper, a set of test cases was constructed based on the real industrial applicationand the methods proposed in the published literature, and extensive numerical experiments were conducted. The experimental results show that, firstly, compared to the Gurobi solver, the algorithm designed in this paper demonstrates significant advantages in solving efficiency and can obtain high-quality solutions in a short period of time. In addition, based on a large number of numerical experiments, it was found that the lower bound of the model has a significant acceleration effect on the first stage problem by cutting and reducing the size of variables, and the average solving efficiency is improved by nearly 200-300 times without affecting the quality of the solution; The dynamic programming problem in the second stage has a significant effect on eliminating model symmetry, which is reflected in a significant reduction in the solving time; Each module of the iterative optimization model make a significant contribution in improving solution quality.