Arsh Zahed
Papers
1
Total Citations
12
H-Index
1
About
Arsh Zahed’s research lies at the intersection of robotics, imitation learning, and human-robot interaction, with a focus on developing algorithms that enable robots to learn complex behaviors from dynamic, real-world supervisors. In his seminal work, “On-Policy Robot Imitation Learning from a Converging Supervisor” (2019, 12 citations), Zahed addresses a critical gap in imitation learning: existing methods like DAgger assume a fixed, unchanging supervisor, which fails in scenarios where the supervisor—whether a human mastering a novel task or an improving algorithmic controller—evolves during policy learning. By formalizing and solving this problem, he provides a framework for robots to robustly learn from non-stationary experts, significantly advancing the practicality of imitation learning in dynamic environments. This contribution has implications for adaptive robotics, where agents must continuously refine their skills alongside human collaborators. Zahed’s work is notable for bridging theoretical rigor with real-world applicability, earning recognition for its potential to enhance autonomous systems in manufacturing, healthcare, and service robotics. His research continues to inspire new approaches to robot learning from imperfect, evolving sources.
Research Focus
Key Achievements
Top Papers
- 1On-Policy Robot Imitation Learning from a Converging Supervisor12 citations · 2019