============================================================================== Department of Computer Sciences The University of Texas at Austin Austin, TX USA The TacTex Starter Agent (for the Trading Agent Competition - Supply Chain Management scenario) Copyright 2006 David Pardoe and Peter Stone. All rights reserved. You are free to use, modify, and redistribute this software for noncommercial purposes so long as you acknowledge its use in any publications or releases of code or binaries. This software is provided "as is" and without warranty of any kind. This release includes the TAC SCM Java AgentWare 0.9.6 Beta (scmaw.jar) and the TAC SCM Game Data Toolkit 0.4.1 beta (scmlogtool.jar), both developed by the Swedish Institute of Computer Science. The latest versions, including source code, may be downloaded from the TAC homepage (www.sics.se/tac). Citation: to cite use of this agent, we request that you cite the TacTex agent, from which it is derived: David Pardoe and Peter Stone. TacTex-05: A Champion Supply Chain Management Agent. In Proceedings of the Twenty-First National Conference on Artificial Intelligence, pp. 1489–94, July 2006. Here is a bibtex entry: @InProceedings{AAAI06-tac, author="David Pardoe and Peter Stone", title="{T}ac{T}ex-05: A Champion Supply Chain Management Agent", booktitle="Proceedings of the Twenty-First National Conference on Artificial Intelligence", month="July", year="2006", pages="1489--94", url="http://www.cs.utexas.edu/~dpardoe/papers/AAAI06.pdf" } ============================================================================== The purpose of this agent is to serve as a starting point for new participants in the TAC SCM competition. The agent is based on TacTex, winner of the 2005 and 2006 competitions. The basic infrastructure of TacTex has been left in place, but all agent strategies have been replaced with simple heuristics. (More information on TacTex is available at http://www.cs.utexas.edu/~TacTex) The agent is designed for the 2006 TAC SCM competition rules and the corresponding server (version 0.8.18 at the time of the competition), although it will likely remain compatible with any server updates in the near future. All files needed to run the agent are included. Currently available servers include tac3.sics.se:8080 through tac6.sics.se:8080, or you can download and run your own server. For information on the rules, servers, or TAC in general, please see the TAC homepage: www.sics.se/tac The agent will first need to be compiled using compile.sh To run the agent: java -cp scmaw.jar:. se.sics.tasim.aw.client.Main -config .conf (Note: Windows users should replace the ":" with ";".) When the agent is run, the server will automatically create a new game. There is no need to manually start and join a game on the server. An agent requires TWO different configuration files to run. The first is the .conf file. A Sample.conf file has been provided - this file will need to be modified with the desired agent name, password, and server; just replace the current values of NAME, PASSWORD, and SERVER. The second configuration file will be read by the agent at the start of each game to set agent parameters such as the classes to use for various tasks. The name of this file is Config.txt, where is the complete name of the agent (e.g., Agent1 and Agent2 would need separate configuration files). The SampleConfig.txt file provided will need to be renamed before the agent is run. The reason for this naming system is that no parameters can be passed to the agent, so the agent name is used in place of parameters. By parsing the game logs provided after each game, is is possible to restore the state of the agent on any game day and run the agent for that day. This is very useful for debugging the agent. To run the agent named on day using a particular (a slg.gz file) for a game that it played in: java -cp scmaw.jar:scmlogtool.jar:. TACAgentParserMain -handler TACAgentParserHandler -file -agentName -date The easiest way to run the agent multiple times is with a simple script such as run.sh The agent is released "as-is" with no support promised. However, we request that you let us know if you are using it for your research purposes, and we will do our best to respond to any problems that may arise when trying to run it. More information about the agent is available at: www.cs.utexas.edu/~TacTex/starterAgent To contact us, email: dpardoe@cs.utexas.edu.