Papers
12
Total Citations
223
H-Index
6
About
Nitin J. Sanket is a robotics and computer vision researcher whose work sits at the intersection of aerial autonomy, perception, and resource-constrained intelligence. His research focuses on enabling small aerial robots — particularly quadrotors and nano-UAVs — to navigate complex environments using minimal sensing and computation, a challenge he approaches with both algorithmic ingenuity and hardware awareness. Sanket's most influential contribution, *GapFlyt* (2018, 90 citations), introduced an active vision framework allowing quadrotors to detect and traverse narrow gaps without building explicit environmental maps — a landmark step toward truly autonomous aerial navigation. His *SalientDSO* work (2019, 73 citations) advanced visual odometry by incorporating semantic saliency into sparse direct methods, improving localization in cluttered indoor scenes. Across his PRGFlow series, he tackled the critical challenge of developing deep optical flow and visual-inertial odometry systems that respect the strict size, weight, and power (SWAP) constraints of aerial platforms. More recently, Sanket has pushed boundaries in uncertainty-aware perception (*Ajna*), zero-shot object segmentation (*NudgeSeg*), generative simulation (*WorldGen*), and ultra-efficient optical flow for tiny robots (*EdgeFlowNet*). His body of work reflects a rare commitment to building perception systems that are not merely accurate, but deployable on the smallest, most parsimonious robotic platforms imaginable.
Research Focus
Key Achievements
Top Papers
- 1
- 2SalientDSO: Bringing Attention to Direct Sparse Odometry73 citations · 2019
- 3
- 4
- 5NudgeSeg: Zero-Shot Object Segmentation by Repeated Physical Interaction7 citations · 2021
- 6PRGFlow: Benchmarking SWAP-Aware Unified Deep Visual Inertial Odometry6 citations · 2020
- 7
- 8<i>EdgeFlowNet:</i> 100FPS@1W Dense Optical Flow for Tiny Mobile Robots5 citations · 2024
- 9WorldGen: A Large Scale Generative Simulator5 citations · 2023
- 10Active Vision Based Embodied-AI Design For Nano-UAV Autonomy3 citations · 2021