불확실성 정보를 활용한 Stochastic ZEM/ZEV 유도법칙

Abstract

In this study, we construct a neural network that can optimize the position of cargos called Unit Load Devices (ULDS) on multiple consecutive flights by applying the attention model and multi-agent deep reinforcement learning. The arrangement of ULDs in the cargo plane causes a change in the stability of the cargo plane, fuel efficiency, and unnecessary unloading at the intermediate point. Accordingly, the goal of this problem is to ensure the stability of the cargo plane and minimize the cost incurred. To solve this problem, we try to solve this problem by applying an attention model that shows good performance in combination optimization. The proposed model consists of three encoders and two decoders, and the model is trained through the REINFORCE algorithm. We validate the learned neural network through case study. It was confirmed that the sufficiently learned model can derive a quasi-optimal solution within a short computational time.