Damn it’s hot today…
Some basic notes to start with.
Default piece orientation
- I-piece: [ ][ ][ ][ ]
- O-piece: Self-explanatory…
- T-piece:
-|(Vertical with left point). - Z-piece: Z.
- S-piece: S.
- J-piece:
¬. Like a not sign. - L-piece:
__|.
Piece hits bottom
From observation, it is known when the piece hits the bottom when prevObservation == o or when the drop action is done. For the drop, a new piece comes in the step after the action is done. For the other, it takes one step longer for a new piece to come in.
Differing MDPs
The different MDPs we have been given for testing appear to only be different in field size and where the pieces spawn (not in the centre everytime).
What was done in the lab today
Started over on the SmartTetris agent now that I have a pseudo overview of the program and can see what needs to be done. Wrote basic code and left commented TODOs for functions that need to be written for proper operation. Also, as stated above, looked more into program operation.
Fleshed out a function and created test case skeletons. All functions should be tested before operation due to the complexity of this task and the errors that could result.