Practical methods of optimization; (2nd ed.)
Practical methods of optimization; (2nd ed.)
Layered Learning in Multiagent Systems: A Winning Approach to Robotic Soccer
Layered Learning in Multiagent Systems: A Winning Approach to Robotic Soccer
Q-Learning in Continuous State and Action Spaces
AI '99 Proceedings of the 12th Australian Joint Conference on Artificial Intelligence: Advanced Topics in Artificial Intelligence
Bicephal Reinforcement Learning
Bicephal Reinforcement Learning
Hi-index | 0.00 |
A desirable elementary behaviour for a robot soccer player is the moving ball interception and shooting behaviour, but generating smooth, fast motion for a mobile robot in a changing environment is a difficult problem. We address this problem by formulating the specifications of this behaviour as a quadratic programming optimisation problem, and by training a neural network controller on the exact solution computed off-line by a quadratic programming problem optimiser. We present experimental results showing the validity of the approach and discuss potential applications of this approach in the context of reinforcement learning.