与传统调度模式不同，协同制造模式下企业之间的调度模式极其复杂。协同企业间的加工工序路线并不固定，且不同类型产品具有不同的加工路线网络。为此本文针对平衡型、瓶颈型、跳跃型、混合型四类具有典型特点的协同制造网络G_p进行分析和设计；考虑制造企业同类产品合并加工策略，构建基于连续加工量的分段生产成本函数；通过设计合理的订单最早交货时间和最晚交货时间，对订单交货进行时间窗口约束，并在此基础上构建了由制造商生产成本W_c^m、订单等待W_s^k（Q_k， T'_k）和提前完工库存成本W_s^k（Q_k， T″_k）、延期惩罚成本构成W_l^k（Q_k， T'″_k）的目标函数。为求解该模型，创新性将蒙特卡洛思想引入蚁群算法，提高蚂蚁选择合理性，避免局部最优；同时，采用移动窗口[min， max]奖励机制，并且对信息素奖励乘以平衡系数k（N）提高奖励可信度，加快搜索速度并提高求解性能。仿真结果表明，本文构建调度模型合理，可以获得优化的调度结果；同时，本文提出的蚁群改进寻优算法具有良好的求解速度和收敛性，算法具有较好的稳定性。

Since the collaborative and cooperative manufacturing mode is gaining popularity, which has the merit of utilizing superior resources in collaborative factories to improve production efficiency, thus it becomes important to do the study on scheduling in collaborative manufacturing mode. In a collaborative manufacturing mode, there are multiple processing paths that forming the collaborative manufacturing networks, and thus the corresponding networks are dynamic and changable. Additionally, different type products belong to different type collaborative manufacturing networks and this makes our model more complex. In light of that, four general types of collaborative manufacturing networks G_p are constructed and discussed, including balance type network, bottleneck type network, jump type network, and hybrid type network. Some pramaters of scheduling model are also designed to make problem more reasable, i.e., production cost function, earliest delivery time t_f^k and latest delivery time t_l^k. An objective function composed of processing costs W_c^m, inventory costs W_s^k(Q_k, T'_k), and the two penalty costs of early completion costs W_s^k(Q_k, T″_k) and tardiness costs W_l^k(Q_k, T'″_k) is then constructed. In order to solve our model, an improved ant colony optimization algorithm is presented, into which the Monte Carlo algorithm is incorporated. In particular, the upper confidence bound z_i is used to guide the ant selection. Meanwhile, a moving window award mechanism[min, max] is also designed to improve award criteria. Given that the expectation window moves frequently with the increase of simulation, and thus the credible level of expectation window increases. In view of this, the fixed pheromone concentration should multiply a balance coefficient k(N) as the award value to improve the rewards credibility. The simulation results show that our designed model is reasonable and practical. Meanwhile, our proposed algorithm has fast solving speed, good convergence and stability. Our research work is benefit for enterprise scheduling in collaborative manufacturing mode.