Abstract
This research is focused on the integration of multi-layer Artificial Neural Network (ANN) and Q-Learning to perform online learning control. In the first learning phase, the agent explores the unknown surroundings and gathers state-action information through the unsupervised Q-Learning algorithm. Second training process involves ANN which utilizes the state-action information gathered in the earlier phase of training samples. During final application of the controller, Q-Learning would be used as primary navigating tool whereas the trained Neural Network will be employed when approximation is needed. MATLAB simulation was developed to verify and the algorithm was validated in real-time using Team AmigoBotTM robot. The results obtained from both simulation and real world experiments are discussed.
Publisher
Trans Tech Publications, Ltd.
Reference12 articles.
1. Elena Garcia, Maria Antonia Jimenez, Pablo Gonzalez de Santos and Manuel Armada, The Evolution of Robotics Research – From Industrial Robotics to Field and Service Robotics, IEEE Robotics and Automation Magazine, Volume 14, Issue 1, Pages 90 – 103, March (2007).
2. Alessandro Saffiotti, Fuzzy Logic in Autonomous Robotics: behavior Coordination, Proceedings of the 6th IEEE International Conference on Fuzzy Systems, Pages 573 – 578, (1997).
3. Soh Chin Yun, S. Parasuraman and V. Ganapathy, Genetic Goal Oriented Path Planning Algorithm for Acute Obstacle Avoidance in Mobile Robot Navigation, The 2010 International Conference on Intelligent Robotics and Applications (ICIRA 2010), China, Pages 624 – 635, 10 – 12 November (2010).
4. C. J. C. H. Watkins, Learning from Delayed Rewards, PhD thesis, King's College, Cambridge, England, (1989).
5. Richard S. Sutton and Andrew G. Barto, Reinforcement Learning : an Introduction, MA : MIT Press, Cambridge, (1998).