Self-Organizing N eurocontrol

Abst"'uct-Self-organizing neural network solutions to control problems are described. Conlpetitive networks �reate spatial filters and geonletry connedion� in a �elf-organizing fa�h­ ion. The goal position, the obstacles and the object under control all create neural activi­ ties through the filters. Spreading activation that discriminates between the controlled ob­ jectl the goal position and the ob�tades is utilized on the internal representation. Lo­ cal self-training method and Hebbian learn­ ing develops the self-organizing control con­ nections. The algorithm provides maneou­ vring capability in unseen scenes.