Container stacking optimization method based on deep reinforcement learning

Abstract

The invention discloses a container stacking optimization method based on deep reinforcement learning, and the method comprises the following steps: designing m environment state variables to represent the stacking state of each stack according to a container stacking sequence and a container lifting priority, and calculating the stacking state of each stack according to the states of n stacks used in the current step and the states of containers to be stacked, calculating to obtain a state matrix S at the current moment; further extracting features in the state matrix S by designing a feature extraction network to obtain a feature matrix T; taking the feature matrix T as input data of a stacking decision network, outputting probability distribution of each stack, and further selecting one stack to stack containers; and training the feature extraction network and the stacking decision network by using a deep reinforcement learning algorithm, evaluating output by using a decision evaluation network during training, optimizing a stacking decision and updating parameters. The method disclosed by the invention can adapt to the condition that the number of the containers and the highest stacking layer number of the stack are changed, so that the purpose of improving the stacking and extracting efficiency of the containers is achieved.

Description

technical field [0001] The invention relates to the technical field of cargo box stacking, in particular to a cargo box stacking optimization method based on deep reinforcement learning. Background technique [0002] The container stacking problem is a basic decision-making problem that is widely used in logistics, and is often seen in storage systems such as port yards and warehouses. Generally speaking, the storage order of the containers is not completely related to the order of picking up the containers. It is best not to stack the containers that are picked up later on top of the containers that are picked up first. Therefore, if the random stacking method is adopted, it will cause a large number of disorderly stacking, which will cause trouble for the luggage. If the boxes that are picked up at the same time are stacked together, even if a dedicated stack is used, storage space will be wasted. Therefore, there is a need for an optimized method to achieve efficient co...

AI Technical Summary

Problems solved by technology

However, all aspects of the intelligent optimization framework of the deep reinforcement learning method must be redesigned according to the characteristics of the problem, and the existing successfully applied methods cannot be directly applied.

Images