Agent intellectualization of Controlled Objects

Abstract

During learning, tasks done in laboratory makes our knowledge deeper, but not always these tasks can be done. Sometimes we don’t have the equipment or we are just short in time. For those, who are learning in distance, course laboratory tasks are almost impossible. To solve these problems the electronic laboratories are created, which lets to do laboratory tasks through the internet. The problem is a insufficient adaptability of the most part of e. laboratories in operation. E.laboratory system which is being created in Klaipėda University will be used for training how to control the object. Student, after choosing the object, will have to write some scripts to control the object, the system, after the laboratory task is finished with the real robot, will inform about the results. For script evaluation training algorithms will be used, one of them is Q-learning algorithm. To make the system more independent, it will have some agents. Agents will coordinate and administrate the system. Also agents will make there own laboratory tasks, to determine the most optimal object control scripts. The whole system can be interpreted as one powerful self learning agent, which helps students to learn and learns itself, to achieve his intention: to control the object as good as possible. Ill.5, bibl.11 (in Lithuanian; summaries in Lithuanian, English and Russian).