Ad hoc teamwork is a relatively new research area that examines how an agent ought to act when placed on a team with other agents such that there was no prior opportunity to coordinate behaviors. In some team domains, such as search and rescue missions and many team sports, the team behavior can be broken down into roles. As such, I have presented a role-based approach for ad hoc teamwork, in which each teammate is inferred to be following a specialized role that accomplishes a specific task or exhibits a particular behavior. In such cases, the role an ad hoc agent should select depends both on its own capabilities and on the roles currently selected by the other team members.
Austin Villa Standard Platform Robot Soccer Team
University of Texas at Austin Learning Agents Research Group
In preparation for Robocup 2012, I worked mainly on creating kicks that could be executed from our walk engine. I also worked on improving our kick from 2011. Before Robocup 2011, I worked to develop a stable, consistent, and powerful kick using a variety of methods. The kick's stability was perhaps its biggest improvement over the previous year. Before Robocup 2010, I worked mainly on two technical challenges - the passing challenge and the dribbling challenge. We placed 1st in the passing challenge and 2nd in the dribbling challenge.
I worked in Ashwin Ram's Cognitive Computing Lab from Summer 2008 to Spring 2009 as part of Georgia Tech's undergraduate research option program. I worked on the Darmok 2 project, which sought to enable a game artificial intelligence to play strategically and learn from experience in real-time strategy games. My work focused mainly on using first order inductive learning to learn rules for representing opponent strategies. Specifically, I used learned rules to perform plan recognition and classify an opponent strategy as one of multiple learned strategies.
I worked with PhD student Zsolt Kira in Georgia Tech's Mobile Robots lab from Fall 2006 to Fall 2008 on a project attempting to establish a shared understanding between two heterogeneous robots in order to facilitate useful communication and knowledge sharing. My contributions include derivation of three methods for establishing a shared context where each method trades off between accuracy and cost and implementation of a novel remove method for the ID5R algorithm to allow for removal of an attribute that the original robot has, but that is not applicable to the robot with which we are sharing data.