The Concept of an Early Warning System for Supporting Air Traffic Control

This article addresses the issue of loss of separation incidents and discusses currently implemented technological solutions designed to minimize the risk of such occurrences. An evaluation of these solutions is conducted, highlighting their key advantages and disadvantages. Additionally, a literature review of proposed new solutions is presented, emphasizing the necessity of introducing a new system to address previously identified shortcomings. This work proposes an early warning system for potential airspace collisions based on an artificial neural network. Drawing from the literature analysis, five fundamental assumptions for an early conflict warning system to support air traffic control are formulated. Each assumption is justified, with some addressing the weaknesses of existing solutions. The contributions of this paper, in relation to previously analyzed works, are as follows: (1) the system does not rely on the dynamics model of a specific aircraft type, (2) the possibility of radar vectoring (vectors to final) is considered, (3) the input data are not limited to the horizontal plane and time differences, (4) the system does not require identifying the most similar historical trajectories to assess minimum separation values and potential conflicts, and (5) the system is expected to perform better in airspace where radar vectoring prevails compared to flight along standard routes. The research methodology is discussed in detail, including the operational environment of the system and the applied algorithms. A feedforward neural network was selected, featuring 32 neurons in the first hidden layer and 16 neurons in the second hidden layer. The training process was conducted using the Levenberg–Marquardt algorithm, chosen for its fast convergence. The presented analyses confirm that the developed system meets the established assumptions.