Neural Q-Learning Algorithms using Webots
This work demonstrates the capability of a Khepera II robot learning specific behaviours (e.g. obstacle avoidance and wall following) by utilizing Neural Q-Learning controllers based on both IR and vision sensors. A highly flexible simulator is developed using Webots for testing the developed controllers before they are validated on the actual robot. For more information, please refer to selected publications
Obstacle avoidance behavior on real and simulated robot
Wall following behavior on real and simulated robot
Selected Publications
Design and Implementation of the Khepera II using Webots Software
Monash University, 2006
Monash Undergraduate Honours Thesis
Design and Implementation of a Simulated Robot using Webots Software
Proceedings of the International Conference on Control, Instrumentation and Mechatronics Engineering, Johor Bahru, Malaysia, 2008, pp. 850-856
Application of Neural Q-Learning on the Khepera-II via Webots Software
Proceedings of the International Conference on Fascinating Advancement in Mechanical Engineering, Tamil Nadu, India, 2008
Utilization of Webots and the Khepera II as a Platform for Neural Q-Learning Controllers
Proceedings of the IEEE Symposium on Industrial Electronics and Applications, Kuala Lumpur, Malaysia, 2009, pp. 783-788
Link: http://dx.doi.org/10.1109/ISIEA.2009.5356361