Airspace’s increasing demand is a current concern without a solution. Different research projects aim to expand its available capacity by improving software performance, for example, concerning trajectory prediction and conflict detection methods. This research contributes to this goal by investigating the effect of different look-ahead time values on a state-based conflict detection method’s performance. A parallel analysis is made concerning the traffic density and meteorological conditions’ influence.

The simulations use actual air traffic data, obtained from the OpenSky database. Corrective data processing is implemented to minimize the noise due to data resolution issues, and time shifting techniques are analyzed and implemented to counteract the human bias natural in real recorded data. The chosen air traffic simulator is the open-source BlueSky Simulator, which integrates the state-based method analyzed. The data is selected considering the traffic density (Eurocontrol database) and the meteorological conditions (ERA5 data from Climate Copernicus) since Light, Medium, and High bins are created. The traffic density bins generation makes use of k-clustering, while the meteorological conditions bins go through a more complex process to identify atmospheric cold fronts.

The performance is obtained for different classification approaches, showing the impact more flexible metrics have on the results. For flexible metrics, the performance of the state-based conflict detection method is higher than for stricter metrics. For the first one mentioned, values higher than 120s look-ahead time are not fruitful, while, for the second one, all look-ahead times are not effective in state-based conflict detection. An analysis focusing on the flight phase showed the performance is better for the cruise phase, raising the effective look-ahead times to 300s and 180s for each approach, respectively.

Concerning the secondary independent variables, firstly, a higher traffic density environment translates to a lower conflict detection performance. Secondly, the meteorological conditions bins’ difference is not enough to withdraw conclusions, even though it follows a similar trend to the traffic density values.