Abstract

We have spent the last few years developing the control and perception technologies that allow Atlas to not just walk, run, jump, dance, and backflip, but also lift, carry, and even throw heavy objects. In this talk, I will first give an overview of the model-predictive controller that powers all of these behaviors, from dynamic locomotion to dexterous bimanual manipulation. Then, I will discuss how Atlas perceives the world around it and how we integrate perception into the robot's behaviors, with a focus on object manipulation tasks. Along the way, I’ll speak a little bit about the path that brought me from Stevens to Boston Dynamics.

About the Speaker

Kevin Doherty is a research engineer with the Atlas team at Boston Dynamics working on state estimation and perception. Prior to joining Boston Dynamics, he completed his Ph.D. at MIT in the MIT-WHOI Joint Program and AeroAstro department. He received his B.E. in Electrical Engineering from Stevens Institute of Technology in 2017 and his S.M. in Aeronautics and Astronautics from MIT in 2019.