Petra Poklukar
Papers
4
Total Citations
58
H-Index
4
About
Petra Poklukar is a leading researcher at the intersection of robotics, reinforcement learning, and computer vision, whose work is pioneering data-efficient methods for robotic manipulation. Her core research focuses on enabling robots to learn complex visuomotor policies with minimal data, a critical challenge for deploying robots in real-world settings. Poklukar’s most significant contribution is the development of the **Latent Space Roadmap (LSR)** , a graph-based planning framework that operates in a low-dimensional latent space. This innovation allows robots to perform visual action planning for both rigid and deformable objects, capturing the system’s global dynamics without requiring high-dimensional state spaces. Her highly cited work, including papers with 25 and 19 citations, also tackles the problem of policy transfer across different robotic platforms through Bayesian meta-learning, ensuring that a policy trained on one robot can be adapted to another with minimal retraining. By combining reinforcement learning with generative models, Poklukar has created frameworks that drastically reduce the data needed for training visuomotor policies, making her a key figure in advancing practical, generalizable robot learning.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4