TIELT

Testbed for Integrating and Evaluating Learning Techniques

TIELT is a software tool that is designed to faciliate the evaluation of decision systems in simulators. Our initial focus is on decision systems that include machine learning components, and on simulators for several types of game engines (e.g., real-time strategy, discrete strategy, role playing, team sports, first-person shooter), with emphasis on those related to military simulators of Computer Generated Forces (CGF). However, TIELT can be used with decision systems other than those that have learning (or learned) components, and can be used with non-gaming simulators. Please visit the "Implementations" section of our forums to find integrations provided by users of TIELT.

TIELT's development began in December of 2003, and its alpha version (v0.5) became available in October of 2004. This first year was primarily dedicated to producing an initial TIELT API and GUI for interfacing decision systems with game engines. This version is in initial use by teams of researchers located throughout the United States, and it will also be used for workshops and competitions to be held at AAAI'05, IJCAI'05, and ICCBR'05. (Links to these will be added as they become available.)

TIELT is available at no cost for all uses, but provides no warranty. Here is a copy of its "as-is" license.

TIELT's development is sponsored by DARPA's Information Processing Technology Office.