Machine Learning for Robots on Earth and in Space
In this talk, I will present our recent progress in developing learning methods that allow robots to acquire and refine a rich set of motor and perceptual skills. The developed algorithms are highly efficient, scale to robots with many degrees of freedom, and can be used to learn challenging tasks ranging from dexterous manipulation of objects, to locomotion, and even basketball. In contrast to the prevalent 'Big Data' approach, we focus on 'Small Data' regimes that can learn from a minimal set of real-world trials. Novel imitation learning and reinforcement learning methods will be discussed, which allow robots to learn new tasks in autonomous and semi-autonomous settings. Finally, I will also describe Bayesian Interaction Primitives - a powerful approach to modeling symbiotic human-robot interaction for physical collaboration tasks.
Bio:
Heni Ben Amor is an Assistant Professor of robotics at Arizona State University. He is the director of the ASU Interactive Robotics Laboratory. Ben Amor received the NSF CAREER Award in 2018, the Fulton Outstanding Assistant Professor Award in 2018, as well as the Daimler-and-Benz Fellowship in 2012. Prior to joining ASU, he was a research scientist at Georgia Tech, a postdoctoral researcher at the Technical University Darmstadt (Germany), and a visiting research scientist in the Intelligent Robotics Lab at the University of Osaka (Japan). His primary research interests lie in the fields of artificial intelligence, machine learning, robotics, and human-robot interaction. Ben Amor received a Ph.D. in computer science from the Technical University Freiberg, focusing on artificial intelligence and machine learning. His dissertation won the overall best dissertation award at the TU Freiberg. He has won numerous best paper awards at major robotics and AI conferences.