**1. Introduction**

Labor shortage and increasing labor cost are serious problems in today's society. With the concept of Industry 5.0, it is imperative to promote industrial transformation and accelerate the automation and intelligent development of equipment in order to reduce the pressure brought by the rapid rise in labor costs, so more and more intelligent equipment is used in factories and storage environments [1–3]. Nowadays, the status of logistics equipment is increasing, and forklifts, as the main force of logistics handling equipment, have been widely used in many fields, such as factories, ports, and warehouses. However, as the requirements of the operating environment continue to increase, the handling equipment can no longer be operated by human hands, especially in special environments, such as high temperature, and hazardous and explosive environments. Along with the development of driverless technology, forklifts are also slowly approaching advanced technologies, such as intelligent identification, wireless transmission, and autonomous navigation and positioning. Intelligent forklifts can enhance the compound ability of forklifts, improve the overall operation level of forklifts, and gradually add more added value. Therefore, intelligent forklifts are the main development direction of forklifts in the future [2]. The operation of an intelligent forklift is quite straightforward; typically, it inserts and picks up pallets at one preset area before travelling to another to dump

**Citation:** Ren, J.; Pan, Y.; Yao, P.; Hu, Y.; Gao, W.; Xue, Z. Deep Learning-Based Intelligent Forklift Cargo Accurate Transfer System. *Sensors* **2022**, *22*, 8437. https:// doi.org/10.3390/s22218437

Academic Editor: Leopoldo Angrisani

Received: 10 October 2022 Accepted: 1 November 2022 Published: 2 November 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

them off, accomplishing a full pallet transfer procedure. However, implementing such a straightforward approach presents numerous specific difficulties:


We propose a deep learning-based intelligent forklift accurate cargo transfer system to address the aforementioned issues, as well as to increase the resilience and accuracy of the system. The system consists of various components with various sensors that cooperate to finish the pallet transfer operation. We specifically use RGB surveillance cameras to check whether there is any pallet that need to be transported at the pallet storage location. Once we determine that there are pallets, we send intelligent forklifts to the area. We then use the RGB-D (depth) camera that comes with the intelligent forklift to calculate the precise position of the pallets relative to the forklift. Finally, we use a high-precision control algorithm to control the forklift. The following three aspects make up the majority of the system features:


In our warehouse, we employ cameras to monitor pallets and intelligent forklifts to insert and remove pallets, as shown in Figure 1.

(a) Pallet monitoring situation (b) Pallet insertion

**Figure 1.** System operation diagram. We have constructed the entire system that is detailed in this paper in the warehouse. One of the eight cameras in the pallet monitoring system, which can monitor the presence of pallets in the storage area and mark them with red boxes when they are found, is illustrated in (**a**). The intelligent forklift arrives at point (**b**), determines the location of the pallets, and then executes the insertion and extraction procedure depicted in the figure after realizing that the pallets need to be moved.

In the rest of the paper, we discuss related work in Section 2, describe our system in Section 3, then present experimental results in Section 4, and give conclusions in Section 5.
