Research On Airport Gate Assignment Based On Genetic Algorithm And Reinforcement Learning

Airport gate assignment is a classic combinatorial optimization problem,which involves the complex constraints of flight types,departure and arrival times,and functional requirements of multiple airlines.Many domestic airports still use manual allocation However,with the increasing number of airport gates,the number of options for gate assignments increases exponentially,which brings huge workload to the airport dispatcher.Therefore,it is extremely urgent to realize the intelligence and convenience of gate assignments.This paper studies and analyzes the overnight-flight gate assignment according to the scientific research project.In order to make full use of resources,the paper proposes a gate pre-assignment model based on the sequence of terminals.According to the business rules,a bidirectional sorting initialization population strategy is proposed.For the problem of early convergence of simple genetic algorithm,a new population diversity index and convergence degree index are put forward to construct an adaptive operator.Based on the adaptive operator,an adaptive genetic algorithm is proposed.Because the gate assigned constraints is very much,the feasible solution space is highly discretized.To find the optimal solution,a gate pre-assignment model of reinforcement learning based on genetic algorithm is put forward on the base of Markov decision process,which includes the definition of action and state,the construction of state-action table and its update process.Considering the problem that different airport staffs have different preferences for the optimization objectives and it is difficult to determine the weight coefficients of optimization objectives,a gate pre-assignment model and its algorithm based on NSGA-Ⅱ algorithm is proposed.The Pareto solution set of gate pre-assignment is illustrated by numerical examples,which provides the more options for the staff’s decision making.To solve the problem of real-time gate assignment,a real-time flight adjustment strategy based on time period is proposed considering the situation of multiple abnormal flights at the same time.Through MATLAB simulation,the above several methods to solve the gate preassignment of flight are compared with manual allocation.In most cases the results of the proposed methods are better than manual solution,and computation time are shorter.Through the comparison with the existing real-time adjustment strategy of gate assignment,the proposed real-time adjustment strategy of gate assignment is better and verify the effectiveness of the proposed real-time adjustment strategy.