Our lab is actively looking for Fall'17/Spring'18/Summer'18 undergrad/grad intern/research assistant, including international applicants, please directly email your information to me if you are interested.

Current Projects

Attention Model for Deep Reinforcement Learning/Deep Imitation Learning
Ongoing research, see publications for more details.

Brain Inspired Deep Localization
Ongoing research.

Modular Reinforcement Learning
Multi-task reinforcement learning.
Plain reinforcement learning algorithms, altough very fascinating, suffer from the problem called "curse of dimensionality", hence do not scale up to many practical problems. Based on previous research results from our lab, I am working on a new reinforcement learning algorithm based on hierarchical modular behavior architecture, referred as Reinforcement Learning Operating System (RLOS) by Professor Ballard. We decompose original problem into modules, train each module independently using reinforcement learning algorithm, and seek an efficient method to combine modules to produce a good global policy. Practical domains to test RLOS are a pacman-like world and autonomous urban driving task.

Virtual Reality
Modeling human attention allocation in virtual driving task.
Urban driving requires multitasking. We design a virtual urban driving environment using WorldViz Vizard, including lanes, cars, pedestrains, traffic signs, etc. We let our participants use Logitect PlaySeat to drive, and use an eye tracker to track their eye movements. The goal is to develop a model to predict human attention allocation in driving tasks.
Relevant previous research of our lab: link.

Logitech Playseat
(Virtual driving environment, NVIS Head-Mounted Display, Logitech PlaySeat)

Robotic Soccer
Robotic Soccer (Standard Platform).
Professor Peter Stone teaches a very good and challenging Autonomous Robotics class. I became intereted in robotic soccer after taking his class. I joined Austin Villa Robotic Soccer Team in Spring 2014. For more information, please see our team website. We won second place at Robocup 2016 in Standard Platform league. Watch our semifinal game vs. Nao-Team HTWK. Watch our final game vs. B-Human.



Participatory Art Museum
Crowd opinion collection, modeling and visualization in semantic space
A project with Edith Xiaoyu Zeng at UTexas Department of Art and Art History, see our project page.