Ph.D. candidate Kevin Haughn first authored an article in the February 2022 issue of the Journal of Intelligent and Robotic Systems. Working under faculty advisor Daniel J. Inman, Harm Buning Collegiate Professor of Aerospace Engineering and co-author of the paper, Haughn’s research focuses on smart materials and machine learning to create intelligent morphing uncrewed aerial vehicle (UAV) systems.
The article “Autonomous Learning in a Pseudo-Episodic Physical Environment” explores how reinforcement learning has proven to be a difficult task outside of simulation when applied directly to physical hardware. By designing a reinforcement learning scheme to autonomously implement traditionally episodic algorithms they showed that the pseudo-episodic technique allows for additional learning updates with off-policy actor-critic and experience replay methods to improve speed and consistency of learning.
