Auto Bagger Process Automation System Utilizing a Humanoid Robot: An Initial Feasibility Study

Auto Baggers are widely used to package non-fragile products such as garments. However, fully automating the process remains challenging because garments and poly bags are both deformable, requiring complex contact-rich manipulation. This paper presents an initial feasibility study of a humanoid-based system that performs the complete Auto Bagger workflow, including grasping a garment package, inserting it into a poly bag, and actuating two safety buttons for sealing. An Action Chunking with Transformers (ACT) policy was trained using 61 demonstrations, with LiPo-based smoothing and quintic interpolation implemented during execution. Experiments on a humanoid robot provided preliminary evidence supporting the feasibility of autonomously completing the sequential packaging process under the tested conditions. The robot completed the entire process in 9 of the 31 evaluation episodes (29.0%). Success rates were 67.7% (21/31) for grasping, 47.6% (10/21) for insertion after successful grasping, and 80.6% (25/31) for button actuation across all episodes. These results highlight the potential of humanoid robots for garment poly-bagging while identifying insertion stability, failure recovery, and safety-aware execution as key challenges for reliable deployment.