The Machine Learning Center at Georgia Tech invites you to a seminar by Aleksandra Faust, a staff research scientist at Google Brain Robotics.
Deep Learning Motion and Task Planning
To complete a task a planning agent must be able to control the robot, understand the abilities and limitations of the control policy, prioritize and select attainable subgoals, and come up with a safe and a feasible plan in a timely manner. In this talk, I will discuss our current progress and the role of the deep learning in each planning phase to learn plans and motions that generalize to unseen real-world environments. First, evolutionary algorithms automate reward design in reinforcement learning and result in end-to-end policies that avoid moving obstacles and transfer from simulation to reality. Those policies incorporate, not only robot space occupancy, but also uncertainties coming from sensors and dynamics for wide classes of robots: differential drive robots with kinodynamic constraints: car, legged robot, etc. with both 1D and 2D depth sensors. Second, deep neural networks learn to estimate the difficulty of the motion to aid the selection of task subgoals, and even to identify important and feasible milestones that the agent needs to reach in order to complete the task. Third, we discuss how curriculum learning and quantization techniques aimed at enabling deep learning to run in real world.
Aleksandra Faust is a Staff Research Scientist at Google Brain Robotics, specializing in robot motion planning and reinforcement learning. Previously, Aleksandra led machine learning efforts for self-driving car planning and controls in Waymo, and was a researcher in Sandia National Laboratories. She earned a Ph.D. in Computer Science at the University of New Mexico (with distinction), and a Master's in Computer Science from the University of Illinois at Urbana-Champaign. Her research interests include machine learning for safe, scalable, and socially-aware motion planning, decision-making, and robot behavior. Aleksandra won the Tom L. Popejoy Award for the best doctoral dissertation at the University of New Mexico in Engineering, Mathematics, and Sciences in the period of 2011-2014, and was named Distinguished Alumna by the University of New Mexico School of Engineering. Her work has been featured in the New York Times, PC Magazine, ZdNet, and was awarded Best Paper in Service Robotics at ICRA 2018 and Best Paper in Reinforcement Learning for Real Life (RL4RL) at ICML.