Papers
14
Total Citations
719
H-Index
8
About
Sulabh Kumra is a robotics researcher whose work sits at the compelling intersection of deep learning, computer vision, and robotic manipulation. His research has primarily focused on enabling robots to intelligently perceive and interact with unknown objects in unstructured environments — a longstanding challenge in autonomous robotics. Kumra's most influential contribution is his pioneering work on robotic grasp detection using deep convolutional neural networks, published in 2017, which has accumulated over 540 citations and helped establish deep learning as a viable framework for robotic grasping tasks. Building on this foundation, he developed the Generative Residual Convolutional Neural Network (GR-ConvNet), capable of generating real-time antipodal grasp predictions from multi-channel scene images — work that continues to draw significant research attention. Beyond grasp detection, Kumra has explored deep reinforcement learning for robotic manipulation, reward shaping for multi-step task learning, and teleoperated humanoid robot systems, demonstrating a broad command of both software intelligence and hardware integration. His early work on 6-DOF robotic arms and omni-directional vehicles reveals a strong mechatronics background underpinning his later AI-driven research. With over 700 cumulative citations, Kumra's contributions have meaningfully shaped how the robotics community approaches data-driven manipulation and autonomous grasping.
Research Focus
Key Achievements
Top Papers
- 1Robotic grasp detection using deep convolutional neural networks543 citations · 2017
- 2
- 3Robotic Grasp Detection using Deep Convolutional Neural Networks30 citations · 2016
- 4
- 5
- 6
- 7Robotic Grasping using Deep Reinforcement Learning8 citations · 2020
- 8
- 9Design and development part 2 of Dexto:Eka: - The humanoid robot5 citations · 2013
- 103D MODELLING AND DESIGNING OF DEXTO:EKA:3 citations · 2012