A Study On Airport Parking Space Allocation Method Integrating Multi-Objective Evaluation Function

With the rapid development of China’s civil aviation industry,the number of flights is constantly increasing,leading to a prominent scarcity of airport resources.The development of resource allocation plan in the operation of airports is directly related to the effectiveness of airports in daily operation.As a major component of airport resources,the allocation and management of parking spaces are crucial in the overall resource scheduling of airports.A good parking space allocation plan can ensure the environmental safety of the airport in the daily production process,reduce the cost of aircraft operation,and save the cost of transit time for passengers during boarding and landing.Therefore,improving the efficiency of parking space allocation plays an important role in enhancing passenger satisfaction,strengthening the cooperation between airports and airline companies,and improving the operational efficiency of airports.In this paper,we combine deep reinforcement learning techniques with real data from the operation of an airport in China based on multi-objective planning to study the allocation of parking spaces.The main research contents are as follows.First,according to the actual operation of an airport in China,the composition of airport parking spaces is analyzed,and the specific parameters of each parking space and its distribution are studied.Based on the operational environment of the airport,the allocation process of the parking spaces,the mandatory constraints to be satisfied by the parking space allocation and the priority principles in the parking space allocation are analyzed.Second,based on the flight history data of the studied airports,the data are firstly preprocessed to ensure the data validity of the subsequent algorithm experiments and no dirty data affect the algorithm solving process.Then,the pre-assignment rules are established by combining the mandatory constraints of the parking space assignment rules.Then,the multiobjective evaluation functions were fused by the settings of four perspective evaluation functions,namely,the shortest distance traveled by passengers,the most efficient utilization of parking spaces,the lowest cost of aircraft taxiing,and the highest utilization of near aircraft spaces,respectively,using the linear weighting method of multi-objective programming.Finally,the parking space allocation model that fuses the multi-objective evaluation functions is constructed.On this basis,the feasibility of the downtime allocation model is verified by combining with the improved particle swarm algorithm.The evaluation value obtained is 4.207,which is 4.63% higher than the manual allocation result.Third,combining the characteristics of deep reinforcement learning,the Actor-Critic algorithm is designed using Mask overlay and multi-threaded asynchronous method for invalid actions and poor convergence in the process of aircraft parking space allocation,and is verified and analyzed.The test results show that the algorithm has good convergence.Last,Using actual airport operation data,the experiments were compared with algorithms related to parking space allocation,and the results showed that the improved multi-threaded asynchronous Actor-Critic algorithm obtained an evaluation value of 4.496.Higher than the manual allocation effect by 11.81%,improved particle swarm algorithm by 7.18%,forbidden search algorithm by 2.41% and immune genetic algorithm by 4.42%.