ADAPTIVE NEURAL NETWORK CONTROL SYSTEM OF AN AUTONOMOUS ROBOTIC VEHICLE BASED ON ONLINE SUPERVISED LEARNING
Abstract
The disadvantages of the classical architectures of artificial neural networks (ANNs) in the problems of intelligent control of an autonomous robotic vehicle are described. Based on updated bi-directional associative ANNs an adaptive neurocontroller has been developed which enables one to find the cause-effect relationships in the “robot-environment” system. The neurocontroller is based on the rule-based system and contains two ANNs that perform two different functions. The first one is implemented as a motoneurons unit that contains the robot motion control algorithm, and the second one is designed to identify in the sensory data sequence new patterns that are added to the first ANN based on the supervised learning scheme.
About the Author
R. A. Prakapovich
United Institute of Informatics Problems of the National Academy of Sciences of Belarus, Minsk
Belarus
Belarus
References
1. Новые методы управления сложными системами. М., 2004.
2. Жданов А. А. Автономный искусственный интеллект. М., 2009.
3. Гаазе-Рапопорт М. Г. От амебы до робота: модели поведения. М., 1987.
4. АмосовН. М. Нейрокомпьютеры и интеллектуальные роботы. Киев, 1991.
5. Рассел С. Искусственный интеллект: современный подход. М., 2006.
6. Хайкин С. Нейронные сети: полный курс. М., 2006.
7. омату С. Нейроуправление и его приложения. М., 2000.
8. Прокопович Г. А. // Информатика. 2009. № 3 (23). С. 68-81.
9. Прокопович Г. А. // Информатика. 2010. № 2 (26). С. 54-65.
10. Прокопович Г. А. // Искусственный интеллект. 2013. № 1. С. 181-190.
11. Прокопович Г. А. // Информатика. 2014. № 4 (44). С. 34-43.
Review
Views: 864
Email this article 