Abstract: In order to ensure the on-time delivery of machine tools produced in the machine tool mixed-flow assembly shop, a machine tool mixed-flow assembly line scheduling optimization method based on improved deep multi-intelligence reinforcement learning is proposed to address the problems of the solution quality of the minimum-delay production scheduling optimization model and the slow training speed. A mixed-flow assembly line scheduling optimization model is constructed with the objective of minimum delay time, and a decentralized and decentralized execution of double deep Q network (DDQN) intelligences is applied to learn the relationship between the production information and scheduling objectives. The framework adopts the strategy of centralized training and decentralized execution and uses the parameter sharing technology, which is able to deal with non-stationary problems in multi-intelligence reinforcement learning. Problems. On this basis, a recurrent neural network is used to manage variable-length state and action representations, giving the intelligences the ability to handle problems of arbitrary size. A global/local reward function is also introduced to solve the reward sparsity problem in the training process. The optimal parameter combinations are identified through ablation experiments. The numerical experimental results show that, compared with the standard test scheme, the present algorithm improves the goal attainment facet by 24.1% to 32.3% over the pre-improvement period, and the training speed is increased by 8.3%.