Deep Reinforcement Learning for Carrier-borne Aircraft Support Operation Scheduling

Abstract

The makespan of support operations of carrier-borne aircraft is a key factor affecting the sortie generation rate. The support operation process involves multiple support resources and operational tasks should satisfy serial and parallel constraint relationships. The effective coordination of these processes can be considered as a multi-resource constrained multi-project scheduling problem (MRCMPSP), which is a complex NP-hard problem. In this paper, a deep reinforcement learning (RL) method is designed to solve the problem, including the image representation of the state, the definition of action mapping, and reward function. Deep convolution neural network and advantage actor-critic algorithm (A2C) are utilized to provide a new solution to the scheduling problem, and experimental results show that the effectiveness of the proposed algorithm.