|Prof. Kyriakos G. Vamvoudakis|
Prof. Kyriakos G. Vamvoudakis is one of three experts recently selected by the Army Research Office to organize and document a much-anticipated workshop on reinforcement learning and reinforcement learning games.
Ultimately, the two-day workshop, Distributed Reinforcement Learning and Reinforcement Learning Games, will help the Army prioritize its own research agenda in this area. In the short-term, it is generating a lot of excitement among prominent academic researchers, like Vamvoudakis and his two co-PI's Prof. Frank Lewis, and Prof. Yan Wan (both from the University of Texas, Arlington). Attendance at the April 12-13 workshop will be by invitation only.
"This is a great opportunity for the most accomplished researchers and thinkers in our field to discuss the direction we are taking in key areas, like reinforcement learning technologies, deep learning, explainable artificial intelligence, and games-in-a-multi-agent setting," said Vamvoudakis.
"It is an honor to organize it, but even more so to be a part of this conversation. There is a lot to explore."
Developing research in this area shows much promise for the military, which is increasingly turning to unmanned aerial vehicles (UAVs) to complete missions that are both high-risk and high-value. Reinforcement learning strategies offer expanded capabilities for maintaining full autonomy in environments where incomplete information is a routine threat.
"It has been shown recently that reinforcement learning in games is playing an important role in many engineering and science fields, including control engineering, artificial intelligence, smart cities, and even the economy," Vamvoudakis added. "We'd like to incorporate learning and adaptation into control strategies, but we know there are still some critical challenges in the application of learning and adaptation methods to solve long-standing engineering and computer science problems."
Vamvoudakis predicted that workshop attendees will be eager to jump into an exploration of those challenges, which include: the curse of dimensionality, optimization in dynamic environments, convergence and performance analysis, security and safety, non-equilibrium settings, online implementation, deep learning, and multi-agent systems.
Kyriakos G. Vamvoudakis currently serves as an assistant professor at the Daniel Guggenheim School of Aerospace Engineering at Georgia Tech. He collaborates closely with centers and labs that align with his research, including Georgia Tech's Institute for Robotics and Intelligent Machines, Decision and Control Laboratory, and Center for Machine Learning. His research interests include learning-based control, game theory, and cyber-physical security.