-
CPEN 502: Architecture for Learning System
-
Robot battle tank development based on reinforcement learning and robocode
- Inputs: the states of the robot
- Outputs: the Q-values of all the actions under the state S
Data | Original Range | Quantization for Look Up Table | Scaling for NN |
---|---|---|---|
my energy | [0, around 100] | 0, 1, 2 | [0, 1] |
distance to enemy | [0, 1000] | 0, 1, 2 | [0, 1] |
enemy's energy | [0, around 100] | 0, 1, 2 | [0, 1] |
distance to center | [0, 1000] | 0, 1, 2 | [0, 1] |
action | NA | 0, 1, 2, 3, 4 | NA |