Abstract: In order to achieve balanced allocation of production capacity among multiple factories in a group enterprise and reduce operating costs, this study investigates an integrated scheduling problem for distributed factory collaborative production and transportation. A multi-agent simulation optimization model with the objectives of minimizing total cost and the makespan is established considering practical production characteristics. An improved NSGA-II algorithm is proposed for solution. The algorithm is based on order splitting, allocation, and processing, where a three-stage encoding structure is designed. Furthermore, simulation-based decoding is embedded to guide the action execution of each agent. A semi-random initialization mechanism is developed based on the characteristics of the problem to improve the quality of the initial solution, including three strategies: the number of split sub-orders, the scale of sub-orders, and factory allocation. In addition, a unique equal-probability insertion and splitting strategy is introduced. A multi-point crossover operator with an adjustment strategy is designed, while two specialized mutation operators are proposed considering the impact of sub-order scales on objectives. Finally, experiments are conducted through a real-word production scheduling case from a yeast group, and the results show that the proposed improved algorithm is effective and superior in solving the integrated scheduling problem.