Nardos Ashenafi - Strengthening Classical Robot Control via Machine Learning

Nardos Ashenafi
Doctoral Candidate
Boise State University

nardosashenafi@u.boisestate.edu

Strengthening Classical Robot Control via Machine Learning

Biography

Born and raised in Ethiopia, I came to the United States to attend higher education. I received my bachelors in Mechanical Engineering from Boise State University and proceeded to attend my PhD in Electrical Engineering with emphasis in control of dynamical systems and robotics. I currently work in the Robot Control Laboratory at Boise State University. I have worked on various robotics projects involving mechanical design, manufacturing, control design and software integration. Over the last few years, I have focused on control design techniques that combine the structure of classical control with the versatility of machine learning techniques.

Abstract

Since the rapid rise in computing capabilities, machine learning techniques have taken the forefront in the field of robot control design. Recent frameworks such as reinforcement learning can deduce viable controllers from a series of interactions with the robotic system. Unfortunately, such data-driven techniques lose the structure and insight achieved by classical control techniques. This talk discusses how we can integrate machine learning frameworks to enhance the capabilities of classical control design and the complex operations we perform in the learning process.