Решение задачи преследования методами нейроэволюции

The pursuit problem with simple motion on a plane is considered in this paper. It is shown that if fitness function depends on a capture tine and target distance and pursuer is being placed onto Archimedian spiral during learning process, then topology neuroevolution method called "NEAT" is able to efficiently solve the problem. Resultant neural network is being compared with two known theoretical solutions via competition series.