An air combat decision-making method based on knowledge and grammar evolution

The problem of decision-making in autonomous air combat is widely studied in domains like military training, UAV operation, computer game intelligence, etc. In this paper, the Grammar Evolution (GE) approach is applied to derive proper tactics strategy in air combat. GE approach finds solutions in the form of structured programs based on evolution principles, and thus being possible to bridge domain knowledge with generic evolution process to form a general approach for decision-making problems. In our work, the basic GE approach is first tested with a problem-related BNF grammar, which regulates the mapping between the genotype and the programs (phenotype). We firstly test some simple situations to prove that GE can search for a right solution. Next, however, when combat with the enemy with the Min-Max strategy, we find it is inefficient to get robust strategy in the dynamic environment of air combat because the derived strategy often cover only a small portion of the whole decision space. To this problem, we propose some improvement measures in the next work.