Online Airline Baggage Packing Based on Hierarchical Tree A2C-Reinforcement Learning Framework

In this study, we propose a hierarchical tree A2C-reinforcement learning framework for online packing of airline baggage. Our approach aims to solve the critical loading algorithm issue and pave the way for automation in the loading process. The framework integrates a hierarchical tree search for processing external environment data with a deep reinforcement learning module designed around the A2C framework. This module incorporates airline baggage constraints, an enhanced reward function, and defined action and state spaces. To refine the packing strategy, the tree search adjusts leaf node selection probabilities based on complete baggage information, while baggage classification utilizes the Family Unity concept. We validate our algorithm's efficacy by contrasting it with BR, LSAH, discrete deep reinforcement learning and Online BPH algorithm, utilizing the established airline baggage dataset. Our experimental findings reveal that our proposed method's average loading filling rate is 71.2%, superior by 8.5% compared to similar selected algorithms, with faster calculation time. We set up an experimental platform and the results affirm the proposed method's efficacy.