Skip to content

Latest commit

 

History

History
15 lines (12 loc) · 815 Bytes

File metadata and controls

15 lines (12 loc) · 815 Bytes

Reinforcement-Learning-On-Robocode

An overview of all the classes:
1, Parent class LUTRobot is the implementation of Q-Learing with intermediate reward.
2, Child class LUTRobotTermialRewardOnly is the implementation of Q-Learing with terminal reward only.
3, Child class LUTRobotWithSarsa is the implementation of Sarsa.
4, Class LUTable is the implementation of lookup table.

The purpose of the implementation:
In this part I implement Reinforcement Learning for my robot tank based on Robocode environment. Before implementation, the following needs to be decided:
1, which actions I would like to support
2, when & how to generate rewards.

Then I capture the contents of a trained LUT to file and this will be later helpful when replacing the LUT component with a neural net.