Authors:
Fabian Reinisch
;
Michael Strohal
and
Peter Stütz
Affiliation:
Institute of Flight Systems, Universität der Bundeswehr Munich, Werner-Heisenberg-Weg 39, 85577 Neubiberg, Germany
Keyword(s):
Computer Generated Forces, Behaviour Modelling, Beyond-Visual-Range, Air Combat, Pilot Training, Behaviour Tree.
Abstract:
Beyond-visual-range (BVR) engagements are getting more and more frequent in modern air combat. One of the key challenges for pilots here is manoeuvre planning, which reflects their decision-making capacity and can determinate success or failure. To ensure pilot training employing virtual BVR air combat simulations yields success, high accuracy levels of the computer-generated forces (CGFs) are essential. To achieve this, it is substantial to not only replicate and simulate the physical properties of the entities to a sufficient degree, but also to provide them with a close-to-human-like behaviour. In this paper, we propose a general concept to tackle these challenges: First, we introduce flight motion dynamic models (aircraft, missiles, chaff) as well as a jammer. Then, we analyse the workflow of a typical beyond-visual-range air combat engagement, separating it into attack, self-defence and decide. Within this context, we introduce Behaviour Trees as a method to model these tasks an
d explain its benefits. Further plans include the verification and validation of the CGF behaviour within future experimental campaigns that consist of human-controlled opponent aircrafts (pilots) flying against the CGFs. Finally, we provide an outlook to future work in where we intent to employ reinforcement learning for tasks containing many degrees of freedom.
(More)