Papers
12
Total Citations
1,076
H-Index
9
About
Zoe McCarthy is a robotics researcher whose work spans robot learning, manipulation, and motion planning, with particular emphasis on making robots more capable and practical in real-world settings. She is perhaps best known for her pioneering work on deep imitation learning using Virtual Reality teleoperation, which demonstrated how consumer-grade VR headsets and hand-tracking hardware could be leveraged to generate high-quality demonstrations for training robot policies directly from raw pixels — a contribution that has garnered over 590 citations and significantly influenced the field of robot skill acquisition. McCarthy has also made notable contributions to deformable object manipulation, developing geometric analyses of equilibrium configurations for flexible wire manipulation that earned over 180 citations and remain foundational references in elastic rod modeling for robotics. Her work on memory-augmented deep neural network policies for partially observed control tasks further highlights her ability to bridge theoretical rigor with practical robotic learning challenges. Additional contributions include belief-space planning for imprecise articulated robots, topological path planning, and uncertainty-aware grasp planning using multi-armed bandit models. Across these diverse threads, McCarthy's research consistently addresses the core challenge of enabling robust, intelligent robot behavior under uncertainty and partial observability.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3Learning deep neural network policies with continuous memory states90 citations · 2016
- 4
- 5
- 6Proving path non-existence using sampling and alpha shapes41 citations · 2012
- 7Multi-armed bandit models for 2D grasp planning with uncertainty37 citations · 2015
- 8
- 9
- 10Learning Deep Neural Network Policies with Continuous Memory States8 citations · 2015