Trevor Darrell

About

Trevor Darrell is a pioneering researcher at the intersection of deep learning, computer vision, and robotics, whose work has profoundly shaped how machines perceive and interact with the physical world. Best known for his groundbreaking contributions to end-to-end learning for robotic control, his landmark papers on deep visuomotor policies — accumulating over 3,000 citations combined — demonstrated that robots could learn complex sensorimotor skills directly from raw visual input, eliminating the need for hand-engineered perception pipelines. This work helped establish end-to-end deep learning as a cornerstone methodology in modern robotics. Darrell's research spans an impressive breadth, from robotic manipulation — including the elegant geometric approach to autonomous laundry folding — to multimodal tactile understanding, where he explored how robots can learn haptic properties of objects through both touch and vision. His investigations into natural language grounding for human-robot interaction further bridged perception and communication, enabling robots to interpret spatial language in real-world settings. More recently, his work on humanoid locomotion via reinforcement learning signals his continued influence on cutting-edge embodied AI. With over 4,000 citations across these diverse contributions, Darrell remains one of the most consequential figures in robotic perception and learning research.

Research Focus

Key Achievements

Top Papers

1. 1
   End-to-end training of deep visuomotor policies

   1,715 citations · 2016
2. 2
   End-to-End Training of Deep Visuomotor Policies

   1,399 citations · 2015
3. 3
   A geometric approach to robotic laundry folding

   252 citations · 2011
4. 4
   Deep learning for tactile understanding from visual and haptic data

   252 citations · 2016
5. 5
   Real-world humanoid locomotion with reinforcement learning

   151 citations · 2024
6. 6
   Grounding spatial relations for human-robot interaction

   146 citations · 2013
7. 7
   Robotic learning of haptic adjectives through physical interaction

   145 citations · 2014
8. 8
   Using robotic exploratory procedures to learn the meaning of haptic adjectives

   107 citations · 2013
9. 9
   Deep Object-Centric Policies for Autonomous Driving

   103 citations · 2019
10. 10
    Head gestures for perceptual interfaces: The role of context in improving recognition

    94 citations · 2007

Key Collaborators

Contact & Links