Dr. Prashant Doshi
Director, THINC Lab
THINC lab is located in the Boyd Graduate Studies Research Center in the southern part of the UGA campus. Currently, more than 10 graduate students including doctoral and masters students are conducting research in the lab. Research conducted in the lab has been funded by grants from the Air Force and Army Research Offices, National Science Foundation, National Institutes of Health and the industry.
Can a robot plan a path to penetrate a patrol by two other robots after simply observing them from a vantage point?
This challenging application domain is motivating PhD students in the THINC lab such as Ken Bogert to develop new methods for inverse reinforcement learning (IRL) – a relatively new area of research that seeks to learn the preferences of an agent from observations of its actions. Researchers are advancing IRL to real-world situations involving multiple experts, occluded trajectories and partly hidden actions with experimentation being conducted on TurtleBots. Other IRL application domains pursued in the lab involve learning driver models for self-driving cars in collaboration with Toyota. [Learn more..]
 Mazen Melibari, Pascal Poupart, Prashant Doshi and George Trimponias, Dynamic Sum Product Networks for Tractable Inference on Sequence Data, accepted in biennial PGM 2016.
 Muthukumaran Chandrasekaran, Adam Eck, Prashant Doshi, and Leenkiat Soh, Individual Planning in Open and Typed Multiagent Systems, accepted in UAI 2016. [paper]
 Mazen Melibari, Pascal Poupart and Prashant Doshi, Sum-Product-Max Networks for Tractable Decision Making, accepted in IJCAI, 2016. [paper]
 Mazen Melibari, Pascal Poupart and Prashant Doshi, Sum-Product-Max Networks for Tractable Decision Making, accepted in AAMAS 2016 (extended abstract).
 Roi Ceren, Prashant Doshi and Bikramjit Banerjee, Reinforcement Learning in Partially Observable Multiagent Settings: Monte Carlo Exploring Policies, accepted in AAMAS 2016 (full paper). [paper] [talk]
 Yifeng Zeng, Prashant Doshi, Yingke Chen, Yinghui Pan, Hua Mao and Muthukumaran Chandrasekaran, Approximating Behavioral Equivalence for Scaling Solutions of I-DIDs, Knowledge and Information Systems (KAIS), Springer, 2015, in press. [article]
 Xia Qu and Prashant Doshi, Individual Planning in Infinite-Horizon Multiagent Settings: Inference, Structure and Scalability, in NIPS 2015. [paper]
[9/17/16] MS student Maulesh Trivedi is the face of international education at UGA.
[9/17/16] Prof. Doshi is the Exhibits Chair of AAMAS 2017.
[5/25/16] PhD candidate Ken Bogert will join UNC Asheville as a tenure-track assistant professor in Fall’16.
[5/25/16] Prof. Doshi is promoted to Full Professor with a unanimous vote at all levels.
[5/7/16] Toyota renews its research contract with the THINC lab for 2016-17 to continue investigating IRL for self-driving cars.
[12/21/15] Enrique Munoz de Cote (INAOE), Long Tran-Thanh (Southampton), Prashant Doshi and Chris Amato (UNH) will jointly organize a workshop on interactions with mixed-agent types at IJCAI 2016.
[12/21/15] Stefano Albrecht of UTA and Prashant Doshi will jointly give a tutorial on type-based multiagent systems at AAAI 2016.