Loading ...
Sorry, an error occurred while loading the content.

Fantasy CWiF

Expand Messages
  • guytipton <guy.tipton@auab.aorcentaf.af.
    OK, there I was, in the Amazon, fighting the giant anaconda, when I had a thought. This being a rare occasion, I immediately set it to paper. I think we all
    Message 1 of 2 , Jan 1, 2003
    • 0 Attachment
      OK, there I was, in the Amazon, fighting the giant anaconda, when I
      had a thought. This being a rare occasion, I immediately set it to
      paper.

      I think we all agree that it is _way_ to difficult to teach an AI to
      play CWiF. But would it be possible to come up with a set of
      heuristics that would allow the computer to play a limited campaign?

      For example, I call the virtual Irwin into the office and say "You
      get 1 GE ARM HQ, 1 IT MECH, 1 IT MOT, 2 IT INF, a FTR and the use of
      2 IT BB and 4 IT CA. Go forth and close the Suez! FM Kesselring will
      ensure that you have supply path back to the Fatherland."

      1. ADG would have to be convinced that they would never be able to
      bring such a beast to market and make a profit from it. Ha-ha.

      2. For this work, there would need to a well documented API to allow
      the AI to give the appropriate orders to the game engine. Much of
      difficult work (like movement rules, supply paths, terrain
      description) is already done in CWiF.

      3. Assume that any such project would use an open source,
      collaborative development effort based on a freeware
      inference/reasoning engine.

      4. I am not a AI programmer (linear algebra, math modeling and
      project manager). Do we have any AI programmers in the WiF community?

      Maj Guy Tipton
      Lead EI Engineer
      AUAB AOC
      612 AOG/EI
    • incywif <b@djuice.com>
      ... I ... to ... It can be done, IHMO. But one shouldn t expect anything more than mediocre AI anytime soon, and I d highly reccomend building the AI for a
      Message 2 of 2 , Jan 2, 2003
      • 0 Attachment
        --- In wifdiscussion@yahoogroups.com, "guytipton <guy.tipton@a...>"
        <guy.tipton@a...> wrote:
        > OK, there I was, in the Amazon, fighting the giant anaconda, when
        I
        > had a thought. This being a rare occasion, I immediately set it to
        > paper.
        >
        > I think we all agree that it is _way_ to difficult to teach an AI
        to
        > play CWiF.

        It can be done, IHMO. But one shouldn't expect anything more than
        mediocre AI anytime soon, and I'd highly reccomend building the AI
        for a specific set of options (which means it'd be pretty useless
        for all but a "standard" set of options)

        >But would it be possible to come up with a set of
        > heuristics that would allow the computer to play a limited
        campaign?
        >
        > For example, I call the virtual Irwin into the office and say "You
        > get 1 GE ARM HQ, 1 IT MECH, 1 IT MOT, 2 IT INF, a FTR and the use
        of
        > 2 IT BB and 4 IT CA. Go forth and close the Suez! FM Kesselring
        will
        > ensure that you have supply path back to the Fatherland."

        The task you describe is way to complex to be achived with any
        degree of success without an advanced, multicomponent AI, able to
        handle a wide range of tasks, ranging from step to step combat
        dispositions to creating, altering, prioritizing and implementing
        long term, abstract plans and goals.

        If you want to build some kind of partial AI, you need to keep your
        task much more simple and more "one-dimentional", like "protect
        given area from stratbombing with x number of FTR". Even so, you
        have little chance of making anything more than a vaugely useful
        partial AI, since activity limits in CWiF pretty much forces any
        decision to be considered in a gamewide scope.

        >
        > 1. ADG would have to be convinced that they would never be able to
        > bring such a beast to market and make a profit from it. Ha-ha.
        >
        > 2. For this work, there would need to a well documented API to
        allow
        > the AI to give the appropriate orders to the game engine. Much of
        > difficult work (like movement rules, supply paths, terrain
        > description) is already done in CWiF.
        >
        > 3. Assume that any such project would use an open source,
        > collaborative development effort based on a freeware
        > inference/reasoning engine.
        >
        > 4. I am not a AI programmer (linear algebra, math modeling and
        > project manager). Do we have any AI programmers in the WiF
        community?

        Yes, me (albeit I have been buzy on other fronts the last 2 1/2
        years. Making AI is NOT exactly a wellpaid profession, compared to
        several obvious alternatives).

        Ingebrigt

        >
        > Maj Guy Tipton
        > Lead EI Engineer
        > AUAB AOC
        > 612 AOG/EI
      Your message has been successfully submitted and would be delivered to recipients shortly.