| In recent years,with the rapid development of China’s civil aviation industry,the number of flight transportation continues to increase,leading to frequent aircraft congestion in some busy large airports.Aircraft congestion can interfere with the normal operation of flights,resulting in flights not arriving on time and causing aircraft delays.Aircraft congestion not only increases the likelihood of conflicts between flights,but also increases the workload of air traffic controllers,affecting the safe operation of flights,and at the same time,it can cause significant economic losses to airlines.Therefore,it is very important to manage the flow of incoming flights and improve the operational efficiency of flights.On this basis,the scheduling of incoming flights has also become a key research object in the industry.This thesis mainly studies and optimizes the model of arrival flight scheduling problem.Firstly,it describes the relevant overview of air traffic flow management,introduces the methods of air traffic flow management,ground waiting procedures,and airport time slots,and describes the fairness of airport time slot allocation.Next,it introduces the approach flight sequencing in air traffic flow management,and describes in detail the concept,principle,and optimization objectives of approach flight sequencing.The optimization objectives include three aspects: airport,controller,and airline.Two functional models,namely,the minimum total loss of flight delay and the maximum fairness between airlines,are modeled,and the relevant definitions and constraints of the model are explained.Then,the commonly used arrival flight sequencing algorithms and heuristic algorithms are introduced.After comparing and analyzing different types of algorithms,it is found that reinforcement learning algorithms can solve the situation where the computational ability of ordinary algorithms lags behind and heuristic algorithms cannot accurately find the global optimal solution.Then it summarizes the basic principles,concepts,and common algorithms of reinforcement learning algorithms.Based on the previously established flight scheduling model,an approach flight scheduling algorithm based on reinforcement learning is established,using the Q-Learning algorithm in reinforcement learning algorithms,and analyzing the algorithm content and operational process in detail.Finally,the environment and runway of Chengdu Shuangliu International Airport were analyzed,and the model algorithm was simulated using the airport’s incoming flight data.Comparing and analyzing the results obtained by the maximum location exchange algorithm and the reinforcement learning algorithm with the results of the first come first service algorithm,it is found that the maximum location exchange algorithm and the reinforcement learning algorithm reduce flight delay losses by 702 yuan and 2106 yuan compared to the first come first service algorithm;Compared to the first come,first served algorithm,the maximum location exchange algorithm and reinforcement learning algorithm improved airline fairness by 23.35% and 64.15%.Therefore,it is proved that reinforcement learning algorithm has the advantages of faster speed and better optimization in solving the arrival flight scheduling problem. |
