Project Description

For highly non-linear systems or variable environments it is challenging and costly to create an accurate mathematical model for autonomously controlled flight. One alternative method is to bypass the need for a model and instead directly learn desirable behaviors of a controller. My research involves using reinforcement learning to create a controller for MAVs that can make intelligent decisions in an unpredictable environment.

Publications

• Junell, J., Knudson, M. and Tumer, K., “Optimizing Sensor / Neuro-Controller Pairings for Effective Navigation,” Artificial Neural Networks in Engineering (ANNIE), pg. 19-26, St. Louis MO, November 2008
• Junell, J. “Adaptive Methods for Robust Commercial Vehicle Control” Master of Science Thesis, Mechanical Engineering, Oregon State University, 2009
• Junell, J. and Tumer, K. “Robust predictive cruise control for commercial vehicles” International Journal of General Systems, 42(7): pp 776-792, 04 July 2013<