Research on grasp planning of automatic loading and unloading with robot twist lock

With the rapid development of modern logistics industry, container loading rate continues to increase, and the degree of automation of container handling is gradually improving. However, the assembly of container lock is still inseparable from manual work. Therefore, based on the requirements of the automation of container handling logistics process, a scheme is proposed to replace manual container lock handling with robots to improve the automation of container loading. In this paper, the industrial robot of ABB Robotics Company is adopted as the research object. ABB robot’s offline programming software RobotStudio has relatively simple functions in container twist lock handling, especially in the two aspects of grasping, handling and assembly. First of all, in terms of grasping and handling torsional locks, it is impossible to flexibly plan the handling path for complex working environment, and changing the robot path is relatively complicated. In terms of the assembly of twisted locks, only fixed mode assembly operations can be carried out, and flexible intelligent assembly cannot be carried out according to site changes. In view of the above problems, this paper mainly through the secondary development of RobotStudio offline programming software and the study of machine learning algorithms, establishes a virtual simulation laboratory, simulates the grasping, handling and intelligent assembly of twist-lock, and verifies the correctness of the developed grasping scheme for container twist-lock handling with the help of ABB robot platform. The secondary development of RobotStudio offline programming software mainly focuses on the setting of target points and path planning. By analyzing the data structure of the target points, mathematical tools such as quaternion are used to simplify the entry process of the target points: through the data storage and arrangement mode of Queue, the target points are arranged one by one, and the path is constructed along with the movement of the robot. And through the program to mark the nature of the target point, we aims to control the robot to complete the grasping, handling, assembly and other operations. The research of machine learning algorithm is mainly used in the assembly process of torsional lock and container corner parts. Combined with the six-dimensional force torque sensor, the machine learning algorithm such as principal component analysis, cluster analysis and hidden Markov model is used to establish a learning model to analyze, organize and learn the sensor feedback data, and judge the mutual position relationship between torsional lock and container corner parts. And with the secondary development program to control the robot to adjust the attitude, we need to complete the intelligent assembly. We need to build a virtual simulation laboratory, introduce special grasping claws, corner pieces and other supporting equipment, simulate the grasping, handling and assembly of twisted locks on the container loading and unloading site, and verify the applicability of the secondary development program and machine algorithm. Finally, on the robot experiment platform, the actual control test of the planned robot path is carried out to verify the correctness of the developed system.