Automated learning by tactical decision systems in air combat


In this paper we propose a self learning approach which utilizes artificial intelligence methods in air combat. It is based on the generation and evaluation of situations and on the subsequent construction of optimal missions. The choice of the evaluation function depends on the partner’s and the opponent’s priorities which are not necessarily known in advance. A learning algorithm is proposed in order to determine these unknown priorities. Furthermore an algorithm is proposed which enables one to decide if the learnt information is sufficient to win. Implementation of these algorithms and their similarities with well known pattern recognition algorithms are outlined. The use of learning algorithms in games involving competing groups of cooperating players (air combats in particular) is new.