Miscellaneous Projects
Knowledge-based Project Planning
Formal Foundations of CBR
Interactive Entertainment Systems (i.e., computer games) and AI
Plan Discovery
Combining HTN Plan Representations and Semantic Web ontologies for knowledge fusion
Domain Independent Case-Based Planning
TIELT Related Projects
Poki Poker-bot TIELT Agent Description
Goal: Integration of Poker Academy Texas Hold’em poker game with Tielt
Details: We integrated Poker Academy Texas Hold’em poker game with
Tielt as a proof of concept that Tielt can be used to control the
AI player in the game.
Download Integration
Download Demo Movie
Author: Hai Hoang
Stratagus-TIELT Integration
Goal: Show that TIELT can control an AI player in a Stratagus network game
Details: The demonstration of TIELT controling an AI player in Stratagus
requires TIELT have game and game interface models for Stratagus, as well as support
within the Stratagus code for the externalization of the internal game events (which
it does not have by default due to its server/client architecture).
Author: Marc Ponsen, Stephen Lee-Urban
Call To Power 2 API and TIELT Integration
Goal: Develop an API for controlling an AI player in a CTP2 game, and
allow for networked clients to use this API.
Details:
We modified the CTP2 codebase to externalize internal game events, and made these changes accessible
via a few API files. Additionally, we implemented a network protocol that allows the API to be controlled
by a TCP/IP client (eg: TIELT with appropriate knowledge base files).
More information, documents, etc.
Author: Chad Hogg, Ushhan Gundevia, Joe Souto, Stephen Lee-Urban
Unreal-bot TIELT Agent Description
Goal: Program bots that can handle all possible actions in UT and purposefully
seek out opponents, goals, and power-ups.
Details: The version of TIELT originally included a UT bot that had rudimentary behavior
as a "proof of concept." This particular project seeks to updated the UT game model to
represent more operators and events, such as sensing damage. The game interface model
is also being updated to provide more actions, such as Jump and Strafe. Finally, the
Agent Description is being revised to provide more tasks for the bot, such as attacking and finding
inventory items. Overall, this should make an admiral straw-man bot for decision systems
to test against in addition to being an excellent demonstration of TIELT features.
Download an implementation of Unreal Tournament (UT) Bots using
the TIELT environment
Author: Megan Vasta
TMKL to HTN Translation
Goal: Translate the task-method-knowledge language variant used by TIELT (TSXML) into
the HTN language used by SHOP.
Details: It has been argued in the past that TMKs are more powerful (expressive) than HTNs
(references forthcoming); by creating a translator that automatically converts from a TMK
representation to a HTN, one demonstrates that HTNs are as expressive as TMKs. With a translator
available researchers will be able to, for instance, utilize TIELT as the interface to a game
and couple that with SHOP in order to fashion plans directly from the underlying TSXML, thereby
using SHOP as the decision system without having to re-create the effort of re-describing the domain.
Finally, by borrowing from the clear semantics of HTNs, we'll be able to contribute a clear
definition of the semantics of TMKs.
Author: Stephen Lee-Urban
Extended Warcraft II representation
Goal: Create a detailed TIELT game model for Warcraft II
Details: A ready made TIELT game model for Warcraft II will enable researchers to easily
pit their decision systems against this popular RTS game.
Author: John Gerace