Papers
24
Total Citations
1,017
H-Index
14
About
Dinesh Jayaraman is a prominent robotics and machine learning researcher whose work sits at the intersection of robot manipulation, tactile sensing, reinforcement learning, and visual representation learning. He is perhaps best known for pioneering the integration of touch and vision in robotic grasping — his 2018 paper "More Than a Feeling" (353 citations) demonstrated how tactile feedback can dramatically improve a robot's ability to grasp and regrasp objects, a fundamental shift from purely vision-based approaches. This thread continued with his influential work on touch-based control using deep predictive models, collectively amassing nearly 150 citations. Jayaraman has also made significant contributions to large-scale robot learning, contributing to the Open X-Embodiment collaboration (119 citations) and the DROID manipulation dataset (108 citations), both of which advance generalizable, cross-platform robotic intelligence. His research extends into reward learning, with Eureka leveraging large language models for automated reward design, and value-implicit pre-training for universal visual representations. Through reproducible benchmarks like REPLAB and work on safety-critical reinforcement learning, Jayaraman consistently bridges theoretical innovation with practical, deployable robotic systems — making him a compelling figure in modern embodied AI research.
Research Focus
Key Achievements
Top Papers
- 1More Than a Feeling: Learning to Grasp and Regrasp Using Vision and Touch353 citations · 2018
- 2Manipulation by Feel: Touch-Based Control with Deep Predictive Models121 citations · 2019
- 3
- 4DROID: A Large-Scale In-The-Wild Robot Manipulation Dataset108 citations · 2024
- 5Eureka: Human-Level Reward Design via Coding Large Language Models48 citations · 2023
- 6
- 7Manipulation by Feel: Touch-Based Control with Deep Predictive Models26 citations · 2019
- 8Cautious Adaptation For Reinforcement Learning in Safety-Critical Settings26 citations · 2020
- 9Time-Agnostic Prediction: Predicting Predictable Video Frames25 citations · 2018
- 10REPLAB: A Reproducible Low-Cost Arm Benchmark for Robotic Learning24 citations · 2019