Research Of Flight Conflict Avoidance In Autonomous Air Traffic Management

With the increasing volume of Civil aviation transportation,Air Traffic Management faces the challenge of airspace high density and low operation efficiency.The Civil Aviation Administration of China has proposed a number of measures to address the problem,including a shift to autonomous air traffic management.At present,there has been a certain research foundation for this model at home and abroad,and it has begun to explore and run.Efficient flight conflict collision avoidance technology is the primary task to ensure the safety facilities of autonomous air traffic management.It is also of great significance to maintain flight order,relieve airspace operating pressure and reduce the load of controllers.In order to explore the operation mode of autonomous air traffic management and study its flight conflict,the main work of this thesis is as follows:Firstly,the operation characteristics of autonomous air traffic management are analyzed,and the characteristics of flight conflicts of autonomous air traffic management are analyzed by comparing with traditional control methods,and the conclusion is drawn that flight conflicts often occur in the short term and are instantaneous and unpredictable.The flight conflict avoidance problem is divided into two aspects: flight conflict detection and flight conflict avoidance.Secondly,according to the assumption that airborne ADS-B IN equipment is adopted to implement "air-air" monitoring in autonomous air traffic management,a flight conflict determination method is designed.Based on the characteristics that determinate detection method is suitable for short-term flight conflicts,flight dead reckoning is carried out considering the influence of wind direction and speed changes on aircraft bias,so as to realize flight conflict detection in autonomous air traffic management.Then,the DDPG algorithm is used to solve the characteristics of continuous problems in autonomous air traffic management.Combined with the characteristics of autonomous air traffic management,the state space,action space,neural network structure and reward function of the agent are designed.The conflict detection method is nested in the state space,and a DDPG flight conflict avoidance model based on high dimensional state action space is constructed.Finally,the OpenScope simulation platform was used to build the training environment of flight conflict avoidance model,and four groups of 100,000 training times were designed for comparative experiments.Different indicators were used to verify the effectiveness of the model,and the training performance of the agent was analyzed.The average success rate of conflict avoidance was above 99%,which was better than the DDPG model designed under the traditional control.The model designed in this thesis can provide a new way of thinking for solving flight conflict problems.Provide a theoretical basis for our country to improve the airspace operating efficiency and explore independent air traffic management technology.