Leslie Pack Kaelbling

Intelligent Robots Redux

Wednesday, May 18, 17:50-18:50, Room A3


The fields of AI and robotics have made great improvements in many individual subfields, including in motion planning, symbolic planning, probabilistic reasoning, perception, and learning. Our goal is to develop an integrated approach to solving very large problems that are hopelessly intractable to solve optimally. We make a number of approximations during planning, including serializing subtasks, factoring distributions, and determining stochastic dynamics, but regain robustness and effectiveness through a continuous state-estimation and replanning process. I will describe our application of these ideas to an end-to-end mobile manipulation system, as well as ideas for current and future work on improving correctness and efficiency through learning.



Bio:  Leslie is the Panasonic Professor of Computer Science and Engineering at MIT.  She has an undergraduate degree in Philosophy and a PhD in Computer Science from Stanford University, and previously held positions at Brown University, Teleos Research, and SRI International.  Her goal is to make intelligent robots: she did some of the earliest work on reinforcement learning and POMDPs in robotics, and is currently focused on integrating geometric, probabilistic, and logical reasoning.
She is the founder of the Journal of Machine Learning Research, a recipient of the IJCAI Computers and Thought award, and a fellow of the AAAI.  She is not a robot.