David L. Roberts
North Carolina State University, Georgia Institute of Technology
Papers
6
Total Citations
258
H-Index
5
About
David L. Roberts is a leading researcher in human-robot interaction and interactive machine learning, with a focus on enabling non-expert users to intuitively train autonomous agents. His work centers on developing algorithms that allow robots and virtual agents to learn complex behaviors from natural, human-provided feedback—such as praise, punishment, and implicit cues—rather than requiring programming expertise. His most influential paper, "Interactive Learning from Policy-Dependent Human Feedback" (108 citations), fundamentally reframes how agents interpret evaluative feedback by accounting for the teacher’s policy, a breakthrough that has shaped subsequent research in interactive reinforcement learning. Roberts also pioneered the concept of Cyber-Enhanced Working Dogs (CEWD) for search and rescue, integrating canine intelligence with cyber-physical systems to create a novel human-animal-robot team (78 citations). His work on adapting agent action speed to improve task learning from non-experts (33 citations) and leveraging implicit human feedback strategies (25 citations) further demonstrates his commitment to making robot training accessible. With over 250 total citations, Roberts’ contributions are foundational to creating robots that learn seamlessly from everyday human interaction.
Research Focus
Key Achievements
Top Papers
- 1Interactive Learning from Policy-Dependent Human Feedback108 citations · 2017
- 2Toward Cyber-Enhanced Working Dogs for Search and Rescue78 citations · 2014
- 3
- 4
- 5Training an Agent to Ground Commands with Reward and Punishment12 citations · 2014
- 6Optimization problems involving collections of dependent objects2 citations · 2008