Fix up the terminal states to check for death
Merge remote-tracking branch 'seh/master'
Refactor enemy distances on sides
Rearrange existing code to not calc maze distance twice
Shuffle possible actions to help minimax efficiency
Change the text output to debug more easily
Make iterative deepening the default.
Make lastMoveCount a flag.
Add various comments.
Fix strategy switching to avoid randomly switching at the beginning (…
…when everyone is in the same area).
Add feature to prefer not to go to recently visited spaces.
Tweak the weights a bit.
Use Python's inf indicators
Fix logging levels.
Added switching between weight strategies (offense/defense) based on …
…distance to opponent's food.
Modify gitignore so upgradedAgents.pyc is not removed again
Only try maxPlys and nothing less; save time
Move stop feature so both agents can use it
Fix leqInf function
Remove unused feature and weight calculation
Added move-based minimax:
- uses the most likely belief position as the actual position for opponents
- alpha-beta pruning
- iterative deepening up to a given time
Updated runDebug.sh script.
Rename config file to match new scheme
Also update the runDebug.sh script for the new scheme
Merge branch 'original'
Add in new agents and layouts
Revert "Fix agent path choosing"
The original contest files were fixed, so revert this as well.
This reverts commit 753ac8e.
Revert "Move config files to directories"
The original contest files were updated to fix this.
This reverts commit ae6ea31.
Update contest files from updated zip
Add in q-learning
Add some new features and refine some weights
Make defensive agent use belief distribution
Defensive agent will go toward the enemy agents that it believes is
closest. This lets us win against the default agents.
Add debug script
Only log when debugging
Move config files to directories
It's impossible to run different agents against eachother without
putting them into directories.
Just move the config.py files for now.
Change distribution display to color correctly
If the other team wasn't using our agent, the distribution belief array
would be off and it would color the agents incorrectly.
Fix agent path choosing
Refactor belief distribution code
Added all group members to the config