Research On The ATC Conflict Identification And Resolution Based On Machine Learning

Air traffic conflict identification and deployment technology is an important issue in air traffic control.Studying the air traffic conflict identification and resolution technology and developing efficient and practical conflict assistant decision-making tools can effectively reduce the workload of controllers and improve the service support capability of air traffic control units.This paper studies the common control conflict identification and resolution theory,improves the conflict identification algorithm to improve the speed,proposes the conflict allocation method considering the deployment timing and practical instructions,and develops the corresponding program of conflict identification and resolution.In the past,the research on conflict identification and resolution technology mainly focused on the construction of mathematical models.The optimization algorithm was used to solve the theoretical optimal resolution scheme,and the practical application effect of the method was less concerned,resulting in poor practicality of the technology.The identification and deployment of regulatory conflicts involves many aspects of air traffic operations.On the one hand,assistant decision-making tools need to ensure effectiveness,on the other hand,the auxiliary decision-making proposed must have practical value.It is not enough to simply pursue theoretical optimal results.Therefore,this thesis studies and improves the identification and deployment of regulatory conflicts from the perspective of technical practicability.The main research contents are as follows:(1)Research on conflict detection techniques.The conflict identification consists of two parts,the aircraft position prediction after a certain period of time and the conflict judgment of all known aircraft positions.The actual demand for the control conflict identification tool is accuracy and efficiency.The accuracy of the identification depends on the accuracy of the track prediction.This paper develops the aircraft trajectory estimation program based on the BADA aircraft basic database,which can accurately predict the aircraft in the given In order to improve the speed of conflict judgment,this paper proposes a collision detection algorithm based on spatial query technology.The algorithm first constructs the aircraft position data into a spatial data structure for filtering query and obtains conflict candidates.The set,and then combined with the control conflict type judgment algorithm to judge all conflict pairs,and obtain the conflict set.In this paper,the R-tree is selected as the data structure of spatial query.Based on the classic R-tree implementation,the R-tree single-point insertion deletion method and batch insertion method are overridden for improving the performance.Simulation experiments show that the collision recognition module constructed in this paper can accurately predict the trajectory of aircraft in a certain period of time,and the detection accuracy and detection speed of this module for the control conflict are significantly improved compared with some existing research results(2)Research on conflict resolution techniques.The traditional conflict deployment method is directly deployed according to the current position and state of the two aircrafts after the conflict is detected.The relief scheme given has a poor adaptability to the situation change of the sector,and the timing of deployment is not optimal.In this paper,Markov decision process is used to model the allocation process of control conflicts,and Monte Carlo tree search algorithm is used to solve the conflict allocation problem.Within Monte Carlo tree search algorithm,in order to improve the running speed of the iterative process,this paper constructs the nominal speed track prediction model,based on aircraft by precomputation different models in different flight level flight corresponds to the nominal speed to build database,to simplify the aircraft trajectory calculation into a simple motion model calculation,it has been verified by experiment that the simplified track prediction accuracy loss results in a relatively short time in the acceptable range.Based on the laboratory development simulation system of air traffic operation,developed which has the function of conflict identification and allocate controllers Agent components,and the corresponding simulation experiment using the simulation system design,constructed the corresponding conflict scene test,through the analysis of simulation results,verify the conflict identification and deployment algorithm development.Compared with the traditional conflict allocation algorithm,the conflict allocation algorithm based on Monte Carlo tree search can better perceive the change of environment and better grasp the optimal timing of allocation.