Warehouse Robot Path Planning Technology Based on Deep Learning

Abstract

The society’s demand for express delivery storage is growing, the requirements for the automation and intelligence of goods allocation have significantly increased. Warehouse robots based on traditional path planning algorithms have drawbacks such as long planning times and lengthy paths in the context of the current large-scale and complex logistics warehousing environment. This paper aims to sort out the development process of various deep learning-based path planning methods in warehouse robots. Systematically classify those into three types: value function algorithms, policy algorithms, and value-policy hybrid algorithms, and analyze their principles. Then, by discussing the research results of various algorithms, it demonstrates the applicability of deep learning-based path planning in modern warehouse applications. Next, the advantages and disadvantages of various algorithms in the development process are compared in the form of a list, reflecting recent achievements in path planning and pointing out the future prospects and development directions in this field.