Research on adaptive multi-objective dynamic scheduling algorithm considering order uncertainty and intelligent transportation supply chain

The rapid development of intelligent optimization algorithms makes the supply chain scheduling to more intelligent transportation direction, but which still has the problem of insufficient adaptive and multi-objective trade-offs because of order uncertainty. Therefore, the research uses reinforcement learning as a framework to construct the scheduling problem in the supply chain as a mixed-constraint decision-making process through Markov making, and builds an adaptive multi-objective dynamic scheduling algorithm by combining with the multihead attention mechanism. In the subsequent performance test experiments, the computational time consumed by the proposed algorithm in state encoding, policy network inference, Q-value computation of multi-objective and constraint security layer are 40ms, 21ms, 25ms, 20ms respectively, and its convergence rate score is 83 when the number of episodes is ten times, and its task completion rate reaches 93% after a short adjustment during perturbation test, and the mean value of task completion time is 1920min, all of which are significantly better than the comparison algorithm. The above results show that the proposed algorithm has high adaptability, robustness and better efficiency in supply chain intelligent transportation scheduling with order uncertainty, and it is expected to provide new ideas for the development of adaptive multi-objective dynamic scheduling algorithms.