Ilya Kostrikov

About

Ilya Kostrikov is a leading researcher in deep reinforcement learning (RL) for real-world robotics, with a focus on sample-efficient, model-free control. His most impactful work, "Demonstrating A Walk in the Park" (2023, 48 citations), shows that a quadruped robot can learn to walk on diverse outdoor terrains—including grass, mulch, and hiking trails—in just 20 minutes of real-world training, a dramatic leap in efficiency. Kostrikov’s broader contributions include developing FastRLAP (2023), a system that enables an RC car to learn high-speed, aggressive driving from visual observations entirely through autonomous real-world practice, without simulation or human intervention. He also advanced offline RL for visual navigation (2022), allowing robots to optimize user-specified preferences like lane-following or avoiding grass. With over 86 total citations across his top papers, Kostrikov’s work bridges the gap between RL theory and practical deployment, demonstrating that deep RL can be both fast and robust in uncontrolled environments. His achievements highlight a commitment to making RL accessible for real-world applications, from legged locomotion to autonomous driving.

Research Focus

Key Achievements

Top Papers

1. 1
   Demonstrating A Walk in the Park: Learning to Walk in 20 Minutes With Model-Free Reinforcement Learning

   48 citations · 2023
2. 2
   A Walk in the Park: Learning to Walk in 20 Minutes With Model-Free Reinforcement Learning

   25 citations · 2022
3. 3
   Offline Reinforcement Learning for Visual Navigation

   6 citations · 2022
4. 4
   FastRLAP: A System for Learning High-Speed Driving via Deep RL and Autonomous Practicing

   5 citations · 2023
5. 5
   D5RL: Diverse Datasets for Data-Driven Deep Reinforcement Learning

   2 citations · 2024

Key Collaborators

Contact & Links