Ramtin Pedarsani

Papers

Total Citations

H-Index

About

Ramtin Pedarsani is a rising researcher at the intersection of robotics, imitation learning, and autonomous driving. His work focuses on developing more intelligent and generalizable agents that can learn complex behaviors from imperfect, real-world data. A key contribution is his research on heterogeneity in imitation learning, where he addresses the critical flaw of treating all human demonstrators as equally expert. His 2022 paper, "Imitation Learning by Estimating Expertise of Demonstrators" (8 citations), proposes a framework to infer and weight the varying expertise of multiple data sources, allowing an agent to learn more robust policies by ignoring poor demonstrations. This work has direct implications for training safer autonomous systems. Additionally, Pedarsani is pushing the boundaries of generalization in autonomous driving. In his 2021 work, "Towards Learning Generalizable Driving Policies from Restricted Latent Representations" (3 citations), he investigates how constraining learned representations can prevent overfitting to specific driving environments, enabling a policy to adapt more effectively to diverse urban and highway scenarios. Through these contributions, Pedarsani is helping to build the next generation of more capable, data-efficient, and robust autonomous agents.