AI & Fundamentals

Robots that Rapidly Adapt to Diverse Tasks and Environments - Deepak Pathak, Assistant Professor, Carnegie Mellon University

Please register for this event here.

Abstract:

How can we train a robot that can generalize to perform thousands of tasks in thousands of environments? This question underscores the holy grail of robot learning research dominated by learning from demonstrations or reward-based learning. However, it is nearly impossible to supervise an agent whether via demonstrations or rewards for all possible situations it is yet to encounter in the future. Hence, we posit that this generalization ability is only possible if the robot can learn continually and adapt rapidly to new situations. The adaptation has to occur online, at a time scale of fractions of a second, which implies that we have no time to carry out multiple experiments in the physical world and optimizing to estimate various system parameters. In this talk, I will present our early efforts in this direction by decoupling the general goal into two sub-problems: 1) generalization to new environments for the same task and 2) generalization to new tasks in the same environment. I will then discuss how these sub-problems can be combined to build a framework for general-purpose embodied intelligence. The talk will include results from case studies of real-world robots including robots walking in unseen diverse terrains in the real world, generalizing to a range of unseen diverse manipulation tasks in a zero-shot manner, and perform dynamic manipulation tasks like writing digits on white-board, scooping, etc.
 

Bio:

Deepak Pathak is a faculty in the School of Computer Science at Carnegie Mellon University. He received his Ph.D. from UC Berkeley and his research spans computer vision, machine learning, and robotics. He is a recipient of the faculty awards from Google, Sony, GoodAI, and graduate fellowship awards from Facebook, NVIDIA, Snapchat. His research has been featured in popular press outlets, including The Economist, The Wall Street Journal, Quanta Magazine, Washington Post, Wired, and MIT Technology Review. Deepak received his Bachelor's from IIT Kanpur with a Gold Medal in Computer Science. He co-founded VisageMap Inc. later acquired by FaceFirst Inc.

For details: https://www.cs.cmu.edu/~dpathak/ 

Please register for this event here.