Dear colleagues,
The Applied Reinforcement Learning (ARL) Seminar is an online seminar that presents the latest advances in reinforcement learning applications and theory, organized by Drs. Rui Song, Hongtu Zhu, Tony Qin, Jieping Ye and Michael R. Kosorok.
Reinforcement Learning (RL) learns how the agents should take actions when interacting with the environment to obtain the highest reward. Due to many successful applications in robotics, games, precision health, e-commerce and ride-sharing industries, Reinforcement Learning (RL) has gained great popularity among various scientific fields. The goal of the ARL Seminar is to bring you a virtual seminar (approximately) featuring the latest work in applying reinforcement learning methods in many exciting areas (e.g., health sciences, or two-sided markets).
This time, the ARL Seminar is excited to welcome *Liam Paull* from l'Université de Montréal to talk about "Training Robotics in Simulators".
Liam Paull is an assistant professor at l'Université de Montréal and the head of the Montreal Robotics and Embodied AI Lab (REAL), and holds a Canada AI Chair. His lab focuses on robotics problems including building representations of the world (such as for simultaneous localization and mapping), modeling of uncertainty, and building better workflows to teach robotic agents new tasks (such as through simulation or demonstration). Previous to this, Liam was a research scientist at CSAIL MIT where he led the TRI funded autonomous car project. He was also a postdoc in the marine robotics lab at MIT where he worked on SLAM for underwater robots. He obtained his PhD from the University of New Brunswick in 2013 where he worked on robust and adaptive planning for underwater vehicles. He is a co-founder and director of the Duckietown Foundation, which is dedicated to making engaging robotics learning experiences accessible to everyone. The Duckietown class was originally taught at MIT but now the platform is used at numerous institutions worldwide.
The seminar will be on *Wednesday, December 16th, 2020 5:00 PM PT / 7:00 PM CT / 8:00 PM ET / Thursday, December 17th, 9:00 AM Beijing*. Details about the talk can be found on our website <https://arlseminar.com>.
You can get access to the seminar by YouTube Live in the USA or Bilibili Live in China.
YouTube Live Channel: https://www.youtube.com/channel/UCYtw_0jwqtNW0-6NFsPY9BA/live
Bilibili Live: https://live.bilibili.com/22533038
If you are interested in getting updates of our seminar's events in the future, you can register on our website <https://www.arlseminar.com/registration-form/>. We will notify you of the new event at the first time.
We look forward to seeing you.
Best,
ARL Seminar
*Title: Training Robotics in Simulators
*Abstract: Reinforcement learning is an appealing approach to developing robot capabilities. It is flexible and general. However, there are some particular challenges with respect to training RL agents on real physically embodied systems. For example: RL training tends to be quite inefficient and performing rollouts on a real robot system is expensive, real world environments don't automatically reset, and real world environments don't necessarily provide a reward signal to the agent explicitly. To overcome these challenges, training agents in simulators is appealing. However, the new problem becomes ensuring that an agent trained in a simulator generalizes to the real environment, the so-called sim2real problem. In this talk we will present two paradigms for tackling the sim2real, which we refer to as "Learn to Transfer" and "Learn to Generalize". We will also outline some future directions that we are pursuing in the Montreal Robotics and Embodied AI Lab (REAL) in this direction. Finally, I will also briefly describe our AI Driving Olympics project in connection to the problem of robotics benchmarking and "sim2real" transfer.
------------------------------
Zhiwei Qin
Principal Research Scientist, Director of AI Research
DiDi Research America, LLC
Mountain View CA
------------------------------