Abstract: Along with a steady increase in air traffic flow and increasingly severe flight delays, improvement in air traffic management capability is urgently needed on the precondition of maintaining the level of safety. To this end, an evaluation index of air traffic complexity, from three aspects of air traffic flow, flight characteristics and flight conflicts, is proposed in this paper for a certain airspace with controller workload taken into consideration. With flow entering a sector, percentage of aircraft altering altitude, times of speed adjustment, times of changing direction, average time for transiting the sector and number of conflicts taken as the parameters for airspace complexity, a mathematical model for evaluating air traffic complexity is put forward based on L-M neural network algorithm. Through a specific example, the L-M algorithm is compared with traditional algorithm in terms of their calculation results. The results show that the proposed technique possesses a high degree of precision, and therefore, is effective and feasible.

Key words: air traffic management, complexity, neural network, controller workload