Dmitry Kalashnikov

About

Dmitry Kalashnikov is a prominent robotics and machine learning researcher whose work sits at the intersection of deep reinforcement learning, computer vision, and embodied AI. He is best known for pioneering scalable approaches to robotic manipulation, most notably through QT-Opt (2018, 575 citations), which demonstrated that large-scale deep reinforcement learning could enable robots to learn dynamic, vision-based grasping skills. This foundational work laid the groundwork for a series of landmark contributions, including the Robotics Transformer (RT-1, 512 citations) and RT-2 (267 citations), which established how transformer architectures and vision-language models trained on internet-scale data could be transferred directly to real-world robotic control. His collaboration on "Do As I Can, Not As I Say" (2022, 516 citations) further advanced the grounding of large language models in physical robotic affordances. Through the Open X-Embodiment initiative (2023–2024), Kalashnikov has helped build community-wide robotic learning datasets enabling generalist robot models. His cumulative work across MT-Opt and sim-to-real adaptation methods reflects a sustained commitment to making robots more capable, generalizable, and practically deployable, earning him well over 2,000 citations and significant influence in modern robotics research.

Research Focus

Key Achievements

Top Papers

1. 1
   QT-Opt: Scalable Deep Reinforcement Learning for Vision-Based Robotic Manipulation

   575 citations · 2018
2. 2
   Do As I Can, Not As I Say: Grounding Language in Robotic Affordances

   516 citations · 2022
3. 3
   RT-1: Robotics Transformer for Real-World Control at Scale

   512 citations · 2023
4. 4
   RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control

   267 citations · 2023
5. 5
   Open X-Embodiment: Robotic Learning Datasets and RT-X Models : Open X-Embodiment Collaboration⁰

   119 citations · 2024
6. 6
   Open X-Embodiment: Robotic Learning Datasets and RT-X Models

   101 citations · 2023
7. 7
   RT-1: Robotics Transformer for Real-World Control at Scale

   38 citations · 2022
8. 8
   MT-Opt: Continuous Multi-Task Robotic Reinforcement Learning at Scale

   35 citations · 2021
9. 9
   Sim-To-Real via Sim-To-Sim: Data-Efficient Robotic Grasping via Randomized-To-Canonical Adaptation Networks

   25 citations · 2019
10. 10
    Reward Machines for Vision-Based Robotic Manipulation

    18 citations · 2021

Key Collaborators

Contact & Links