BOD / UT

Behavior Oriented Design (BOD) is a way to develop AI systems.  It requires a modular behavior library in any OO language, and a version of POSH action selection for that language.   jyPOSH is a version of POSH that runs in python and jython.  Unreal Tournament is a commercial game that you can buy quite cheaply on ebay, since it's a bit old now.

Installing pyPOSH Unreal Tournament

• First, get a (legal, of course) version of Unreal Tournament.
• We bought our M$ versions off of ebay for about $3 each (be careful of postage charges though.)
• You need to get the classic/original version also packaged as "Best of Unreal Tournament"(or the game-of-the-year edition apparently, though I haven't tried this.)
• Do not buy the 2003 or 2004 versions! (at least, not for this!)  The gamebots interface won't work.
• If you use linux, buy a Windows copy to be legal & then use this to get it working.  Or you can use CodeWeavers.
  
• On Mac OS X, you can use CodeWeavers CrossOver Mac.  It costs money (not much! well worth it), but you can also get a trial demo for a week.
  
• Next, you need to install something called GameBots, which many AI programmers use (that's why we are in UT).
• It's normally easiest to use/install the .umod file.
• But for Mac / CodeWeavers, it seems easier to drag the files from the zip into your UT bottle.
• Start up CodeWeavers, click "manage bottles", then click the button "open C drive in finder"
• Copy the files from the sub directories in the zip directory (once you've unzipped it) into the directories with the same names under drive_c/UnrealTournament
  
• To make it easier to see what's going on with your bot, you might want to use this simplified level file, which is a slight modification of the file distributed with gamebots.
• Move this file to Maps directory under your Unreal Tournament directory.
• Now follow the jyPOSH installation instructions.
• The UT behaviour library comes with our standard jyPOSH distribution.
• You may also want to download ABODE to help edit your POSH plans.
  

• Start UT
• Doubleclick on the game (or run it from the programs menu.)
• If you have any trouble e.g. with graphics, start in safe mode.  You may just want to do this anyway.
  
• When the intro sequence starts, type Esc (the escape button).
  
• To to Game->Start practice session
• "Choose the category "Unreal Tournament"
• Game Type should be "Friendly Remote Bot CTF" (CTF == Capture the Flag)
• Map Name is CTF-SimpleSteve (the one you just copied in.)
• Choose any other settings as you like.
• Click Start.
• Start your BOD bot
  
• Press Alt-Enter to get out of full screen mode of UT (unless you are already in safe mode).
• Type Esc to get your mouse back (suspends UT)
• Run jyposh:
• jython jyposh.py
• This should produce a GUI
  
• Choose the bodbot behaviour library.
• Either:
• Click Choose... button in the Agent Initialisation section.
• Open init_agent for a single bot or init_team for a team of bots
• Click Load Agent Init File button.
• Or:
• Click the Add... button and choose a plan file.
• Click Initalise World and Agent(s) button.
• On the new window, click Start Agent button
• Go back into UT and play!
• press ESC again and go look for your bodbot!
• You may need to switch teams in order to make sure you aren't on the same team as your bot (e.g. if you want to fight it.)
  
• Bots have some trouble seeing waypoints, esp. if they start on the red team, but if you interact with them they can learn by following you.
• You can have two bodbots if you like!
• Just add another plan file using the Add... button.
• Or, like we said above, use the init_team script, or edit your own!
  

Acknowledgements

• Andy Kwong made the initial UT gamebots BOD library (see his 2003 dissertation). 
  
• The current agents were mostly built by Sam Partington (see his 2005 dissertation).
  
• Tristan Caulfield helped with this web page and improved the UT library code & interface (Summer 2005).
  
• Steve Couzens discovered and fixed CMU's hacks on the level file & provided the offense / defense bot "team" plans (Summer 2005). 
  
• Jan Drugowitsch maintained and improved jyPOSH from 2005-2007, including some support of our modifications of the UT code.
• Francis Binns and Andrew Mansfield improved on Sam's code so it deals better with temporally-dependent decisions. They have also been maintaining the UT code as Jan & others improved jyPOSH. (2005-present)