Bart L. Paulhamus

Johns Hopkins University, Johns Hopkins University Applied Physics Laboratory

Papers

Total Citations

H-Index

About

Bart L. Paulhamus is a robotics researcher focused on advancing machine learning for autonomous systems, with key contributions in imitation learning, meta-learning, and task generalization. His work addresses the fundamental challenge of enabling robots to learn and adapt to new tasks with minimal or zero demonstrations—a critical step toward truly intelligent, flexible automation. In his 2022 paper "Visual Goal-Directed Meta-Imitation Learning" (3 citations), Paulhamus explores zero-shot generalization, where policies perform novel tasks without any prior examples, pushing the boundaries of meta-learning efficiency. His companion work, "Learning Generalizable Behaviors from Demonstration" (2 citations), introduces the Primitive Imitation for Control (PICO) framework, which combines imitation learning with task decomposition to create reusable, transferable behaviors. Though early in his citation impact, these papers represent foundational ideas in robotics generalization, tackling the notoriously difficult problem of bridging prior experience to unseen scenarios. Paulhamus’s research is particularly notable for its ambition—aiming to reduce the data and retraining burden in robotic control, which could accelerate real-world deployment in manufacturing, healthcare, and service robotics. His work stands at the intersection of imitation learning and meta-learning, offering promising pathways for building robots that learn like humans: quickly, flexibly, and from limited examples.