Solving a large-scale 3D packing problem with robot constraints

Abstract

Existing algorithms for the container loading problem predominantly focus on manual and pallet packing, with limited research on robotic packing in trucks. This paper develops an Improved Simulated Annealing Algorithm, which is based on “column construction” and satisfies multiple constraints. In addition to general constraints, such as box stability, load capacity, and multiple orders, this study also defines constraints specific to the robot, including continuous box loading, the robot’s working range, and end-effector’s crossover depth. Furthermore, a novel construction method is proposed that organizes boxes into “uniform columns” and “mixed columns,” optimizing random factors and the cooling process within the Simulated Annealing Algorithm, improving the spatial recording method for the “horizontal line.” The experimental results for ISAA promise superiority, achieving state-of-the-art performance in logistics distribution centers and robot packing experiments.