The Design Of General Aviation Airborne Collision Avoidance Logic Based On MDP

After the installation of the Traffic Collision Avoidance System(TCAS)on civil airplane and transport aircraft,mid-air collisions were reduced significantly,however,the general aviation aircraft did not have an anti-collision system.Due to the application of the Automatic dependent surveillance-broadcast(ADS-B)technology in general aviation aircraft,and the reform of low-altitude airspace in China,it is necessary to develop a suitable anti-collision for general aviation.The current TCAS collision avoidance logic defines a number of heuristic rules and mutually coupled parameters.If we want to modify they to accommodate the performance of general aviation,we would do a lot of work and they would be modified extensively.This paper designs anti-collision logic for general aviation based on the Markov Decision Process(MDP),and this method only needs to modify the encounter model and performance indicators,therefore,it can adapt to the rapid development of the aviation industry.This paper introduces the designed method of the general aviation logic unit in two-dimensional space.In order to apply the Markov Decision Process,the dynamic model is discretized into discrete transition model by using sampling and interpolation methods,and the logical table is obtained by using dynamic programming.Then,analyzing the logic table visually and analyzing it by safety curve and evaluating the performance.The main research contents of this paper are as follows.Firstly,the method of extending the collision avoidance logic from two-dimensional space to three-dimensional is studied.In this paper,the encounter problem in three-dimensional space is decomposed into two sub-problems: controllable and uncontrollable,corresponding to the vertical and the horizontal respectively.The dynamic model in the horizontal is derived,and the dynamic programming is used to solve the uncontrollable sub-problem,and the the probability table is obtained.In the encounter simulation,the logic table and the probability table are combined,and the current optimal action can be obtained in real time.The dynamic programming method and the simple point estimation method are compared by visualizing the expected time,encounter simulation,safety curve and Monte Carlo method,and the robustness of the white noise encounter model is analyzed.Secondly,the states are not fully observable due to the sensor noise of the monitoring system.Using the QMDP approximation method,the vertical states are estimated by Kalman filter and the horizontal states are estimated by Unscented Kalman filter.And then the Monte Carlo method is used for performance comparison.Finally,the solution strategy of encounter with multiple intruders is studied.This paper compares closest command arbitration strategy,summation utility fusion strategy and guaranteed cost strategy when there are two intruders in encounter.Then,two utility fusion strategies were compared by simulation and Monte Carlo method.