Garrett Hall
Papers
1
Total Citations
2
H-Index
1
About
Garrett Hall explores the intersection of robotics, machine learning, and security, with a particular focus on the vulnerabilities inherent in imitation-based learning systems. His most cited work, "Studying Adversarial Attacks on Behavioral Cloning Dynamics" (2020), investigates how high-fidelity visual simulations and advanced learning algorithms can train robots through behavior cloning—a method where machines model actions directly from human demonstrations. Hall’s key contribution lies in exposing the fragility of this approach: he demonstrates that even subtle adversarial perturbations can derail cloned behaviors, raising critical safety concerns for real-world deployment. While his citation count is modest, his research is foundational for understanding the security risks in robot learning from demonstration. Hall’s work serves as a cautionary note for the field, urging researchers to build more robust training pipelines. His findings are particularly relevant for autonomous systems in sensitive domains like healthcare or manufacturing, where trust in learned behaviors is paramount.
Research Focus
Key Achievements
Top Papers
- 1Studying Adversarial Attacks on Behavioral Cloning Dynamics2 citations · 2020