Mohi Khansari

About

Mohi Khansari is a robotics and machine learning researcher whose work sits at the intersection of simulation-to-real transfer, robotic manipulation, and scalable robot learning. His research addresses one of the field's most persistent challenges: enabling robots to acquire versatile, generalizable skills without prohibitive real-world data collection. Khansari is perhaps best known for developing RL-CycleGAN (2020, 154 citations), a reinforcement learning-aware image translation framework that bridges the visual gap between simulated and real environments, dramatically improving the practicality of sim-to-real transfer for vision-based grasping. Building on this theme, RetinaGAN (2021, 72 citations) introduced an object-aware approach that further refined domain adaptation fidelity. His work on BC-Z (2022, 89 citations) pushed the frontier of zero-shot task generalization through large-scale imitation learning, demonstrating that robots can tackle novel tasks without task-specific training. Khansari has also contributed to self-supervised object representation learning and long-horizon task planning. Collectively, his portfolio reflects a sustained commitment to making robot learning more data-efficient, scalable, and deployable in unstructured real-world settings — research increasingly vital as robotics moves toward broader practical application.

Research Focus

Key Achievements

Top Papers

1. 1
   RL-CycleGAN: Reinforcement Learning Aware Simulation-to-Real

   154 citations · 2020
2. 2
   BC-Z: Zero-Shot Task Generalization with Robotic Imitation Learning

   89 citations · 2022
3. 3
   RetinaGAN: An Object-aware Approach to Sim-to-Real Transfer

   72 citations · 2021
4. 4
   Data-Efficient Learning for Sim-to-Real Robotic Grasping using Deep Point Cloud Prediction Networks

   29 citations · 2019
5. 5
   Learning Latent Plans from Play

   25 citations · 2019
6. 6
   RL-CycleGAN: Reinforcement Learning Aware Simulation-To-Real

   14 citations · 2020
7. 7
   Online Object Representations with Contrastive Learning

   13 citations · 2019
8. 8
   Modeling Long-horizon Tasks as Sequential Interaction Landscapes

   12 citations · 2020
9. 9
   Online Learning of Object Representations by Appearance Space Feature Alignment

   11 citations · 2020
10. 10
    AW-Opt: Learning Robotic Skills with Imitation and Reinforcement at\n Scale

    10 citations · 2021

Key Collaborators

Contact & Links