Anastasia Varava
Papers
3
Total Citations
28
H-Index
3
About
Anastasia Varava is a leading researcher in robotic manipulation, focusing on enabling robots to plan and execute complex tasks directly from visual input. Her work lies at the intersection of computer vision, machine learning, and robotics, with a particular emphasis on handling deformable objects and high-dimensional state spaces. Varava’s most significant contribution is the development of the **Latent Space Roadmap (LSR)**, a graph-based framework for visual action planning that allows robots to globally reason about manipulation tasks without explicit models of the environment. This work, published in 2022, has already garnered 19 citations for its innovative approach to bridging perception and planning. She has also advanced the field by integrating **Diffusion Policies** and large pre-trained multimodal foundation models to create a robotic skill learning system (2024), enabling robots to acquire new manipulation skills through behavioral cloning. Additionally, her comparative analysis of reconstruction- and contrastive-based models for visual task planning (2022) has provided crucial insights into how robots can learn effective state representations from raw images. Varava’s research is at the forefront of making robots more adaptable and capable in unstructured, real-world settings.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3