Research On Aircraft Cluster Support Problem Based On Reinforcement Learning

The problem of aircraft cluster support is one of the classic problems of combinatorial optimization problems,and there are also a wide range of application scenarios in real life production.For transport airports,efficient fleet support operations can not only save airport operating costs,but also improve passengers’ satisfaction with travel services.For combat aircrafts,the efficiency of higher fleet supporting schedule means that the aircraft’s efficiency is improved which allows the war to gain a greater probability of winning.This paper firstly mathematically models the aircraft cluster support problem,and divides the aircraft’s ground protection process into two stages.The first one is to select a reasonable guarantee parking space,which corresponds to the gate allocation problem.The second stage is project scheduling,which corresponds to the resorces constrained project scheduling problem.In addition,this paper proposes a new solution for the support problem,which disassembles the whole problem into the upper and lower layers of the gate allocation problem and the job scheduling and resource matching problem.In solving the problem of the upper stop allocation,the Q-learning algorithm in reinforcement learning is adopted.In order to improve the efficiency of the solution,the lower layer scheduling problem is considered in the upper layer solution,and the scheduling problem is solved.A greedy algorithm that expects a fast and reliable evaluation of the algorithm.At last,this paper also uses the example designed according to the actual problem to test the proposed aircraft cluster support algorithm based on reinforcement learning,and compares with other algorithms to prove the effectiveness and superiority of the proposed algorithm.