Alexander Pashevich

About

Alexander Pashevich is a leading researcher in robotic manipulation, computer vision, and reinforcement learning, whose work bridges the critical gap between simulation and real-world deployment. His most impactful contribution is the development of methods to augment synthetic images for sim-to-real policy transfer, enabling robots trained in simulation to perform complex tasks in physical environments—a breakthrough that has garnered over 45 citations. Pashevich is also known for advancing hierarchical control in robotics, introducing modulated policy hierarchies and techniques for combining learned primitive skills with reinforcement learning to solve tasks with sparse rewards, such as assembling furniture or preparing meals. His work on plane extraction from depth data using Gaussian mixture regression models further demonstrates his expertise in perception for manipulation. With a citation count exceeding 100 across his most-cited papers, Pashevich’s research is essential reading for anyone interested in scalable, versatile robotic systems that can learn and adapt to dynamic, unstructured environments.

Research Focus

Key Achievements

Top Papers

1. 1
   Learning to Augment Synthetic Images for Sim2Real Policy Transfer

   45 citations · 2019
2. 2
   Learning to combine primitive skills: A step towards versatile robotic manipulation §

   37 citations · 2020
3. 3
   Plane-extraction from depth-data using a Gaussian mixture regression model

   15 citations · 2018
4. 4
   Combining learned skills and reinforcement learning for robotic manipulations.

   5 citations · 2019
5. 5
   Modulated Policy Hierarchies

   5 citations · 2018
6. 6
   Learning to Augment Synthetic Images for Sim2Real Policy Transfer

   2 citations · 2019

Key Collaborators

Contact & Links