K. Niranjan Kumar
Papers
4
Total Citations
34
H-Index
3
About
K. Niranjan Kumar is a roboticist whose research lies at the intersection of interactive perception, manipulation, and reinforcement learning. His work focuses on enabling robots to autonomously explore and interact with complex, cluttered environments. Kumar’s major contributions include developing novel methods for estimating the physical properties of objects, such as mass distribution, through non-prehensile manipulation—a key capability for robots to handle articulated objects like doors or cabinets. He has also pioneered graph-based deep reinforcement learning frameworks that allow robots to intelligently explore structured yet cluttered scenes, such as kitchen pantries or grocery shelves, by leveraging the underlying physical plausibility of the environment. To bridge the sim-to-real gap, Kumar introduced BayRnTune, an adaptive Bayesian domain randomization technique that strategically fine-tunes simulation parameters for more robust real-world transfer. His work on cascaded compositional residual learning further enables complex interactive behaviors by combining high-level planning with low-level motor control. With over 30 citations across his most influential papers, Kumar’s research is shaping the next generation of autonomous robots capable of rich, physical interaction in human-centric spaces.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4