Applications based on artificial intelligence (AI) can be of great benefit in the field of air traffic control, provided it can be proven that such applications can fulfil the stringent safety requirements of the aviation industry. For this purpose, the STARFiSH project will integrate AI-based speech recognition ABSR into an A-SMGCS monitoring system for ground traffic at Fraport AG.
In the project, a joint application of ABSR and A-SMGCS should recognise the instructions spoken by air traffic controllers to pilots, extract the commands (semantics) contained therein and integrate the results into the user interface of the A-SMGCS. An additional safety net for AI applications should thereby ensure that errors in the AI-based speech recognition cannot have a negative impact on the overall system.
The starting point for the safety net is model-based decision-making procedures that provide a framework within which the AI is allowed to act. This is intended to guarantee that even a possibly erroneous behaviour of the AI cannot lead to safety-critical situations. This safety net must withstand the tests that are customary in aviation. If this is successful, nothing will stand in the way of a widespread application of AI in digital assistants and in other safety-critical areas. The findings could subsequently be transferred to other domains in a generalised form.
The application developed in STARFiSH will be prototypically integrated into the simulation environment at Fraport AG. Here, the data for training the AI-based components of the project will also be recorded over the course of several simulations and will be supplemented by recordings from real operations in Frankfurt.
Participants ATRiCS GmbH (Co-ordinator)
DLR Institute of Flight Guidance