Recent Posts

  • R

    @a-s The other AI's don't use UCT or any similar algorithm. They use heuristics based upon testing various possible moves for net TUV benefits and enemy counterattacks.

    read more
  • @rogercooper said in [TestAI] Download and discussion:

    @a-s How did it actually perform?

    Well, technically, it does, what it should and the defined "goals" can be seen in the movement of units on the map.

    In terms of "playing strength", more goals in class "TestSetAndEvaluateGoals" would be needed, and "TestPurchaseAI", as well as "TestPlaceAI" are not elaborated enough, I think.

    "TestPurchaseAI" and "TestPlaceAI" could be connected to the "goal- approach" too, or replaced with other classes for this phases.

    At the moment, they are only connected by an integer "alignmentOfAI" to the rest of the code, and not really bound to it.

    Overall, I would say, it is a "proof of concept", that the UCT algorithm can be applied to the movement phases in TripleA. Of course, it is possible, that this has be done in other AIs of TripleA already.

    read more
  • R

    @a-s How did it actually perform?

    read more
  • About

    "TestAI" for TripleA was originally developed during an university course in 2015 (with TripleA version 1.8.0.9).

    The main goal was to implement an AI for the movement of units based on the UCT algorithm. Other phases of the game are implemented too, but mostly for testing the AI in full game scenarios.

    Download

    Contains all required files and a handbook with further explanations.

    TestAI public.zip

    Development

    If someone is interested in (some parts of) the code, it can be used under the terms of GNU General Public License version 2 or later. I don't have time to maintain the code by myself, but try to answer questions in this forum at least.

    read more