So my problem is with understanding of creation of Q-table for states with more parameters per state like robocode. 99% of all examples online are just too simple and it is hard to imagine it for environment as complicated as this one.
From what I understand in Q-learning is created 2D table of states/actions...so for each game state will be present some action. In game like FrozenLake it is really simple. From each square(X/Y point) you have 4 possible actions(up,down,left,right). It means that in 4x4 FrozenLake field you have 16 possible states and for each state 4 actions(Q-table 16x4). And thats it.
But in robocode you take in mind: turret angle, X/Y coords, Energy and enemies state which contains same parameters. These all parameters represent 1 state, am I right ? So algorithm has to take in mind all of these values. Plus of course all of these values have to be simplified if I want to train it in some reasonable time on my i5-4210H 2.9GHz.
So
1) How to create Q-table for robocode and how will it look like ?
2) What is the best approach to simplify all of these state-values ?
PS: We have it as a school project and we do this in Java/C# because robocode supports both of these languages.