Aditya Patankar

Aditya Patankar

Postdoctoral Associate - Robotics

About Me

I am a Postdoctoral Associate in the Soft Flyers Group at Stony Brook University, New York, where I work on multimodal aerial and underwater robots, along with robotic manipulation and planning. I completed my PhD at the Interacting Robotic Systems Laboratory at Stony Brook, advised by Dr. Nilanjan Chakraborty, with a thesis on Sensor-Based Task-Oriented Grasping. My primary research interests are vision-guided robotic manipulation, task-based grasp synthesis, and motion planning from demonstrations. I envision a future where robots work alongside humans, performing a repertoire of tasks beyond just pick and place operations. My research goal is to develop algorithms and technologies enabling robots to achieve this vision.

During my PhD, I worked on several projects aimed at advancing the state-of-the-art in robotic manipulation, motion planning, and grasp synthesis. I also worked on developing solutions to real-world problems in healthcare (assistive robotics) and agriculture (containerized vertical farming using cobots). Towards the end of my PhD, I explored the potential of utilizing foundation models (VLMs, LLMs and transformer-based architectures) in robotics, particularly for performing complex manipulation tasks, i.e. tasks with constraints on the end-effector’s motion, which may be difficult to specify analytically. Please visit the Research and Publications sections for more information.

During the summer of 2025, I interned at GE Aerospace Research as a Robotics and Embodied AI PhD Fellow in the Autonomous Systems Research group. I evaluated NVIDIA’s robotics stack (cuRobo and FoundationPose) for high-mix low-volume industrial automation and built a perception-driven motion planning framework for autonomously scanning tubes and ducts.

During the summer of 2024, I had the opportunity to intern at Nokia Bell Labs as part of the Modelling and Optimization Group where I was mentored by Dr. Jeongran Lee and Dr. Matthew Andrews. I worked on building 3D representations for dynamic digital twins of indoor environments using radiance field techniques with various camera sensors. This valuable experience enhanced my skills in computer vision, deep learning, and data collection in real-world environments.

Feel free to reach out to me if you are interested in my work, would like to collaborate, or if you need a reviewer for ICRA, IROS, TRO or RAL!

Download CV.

Code

  • tograsp-socp: Python implementation of our task-dependent grasp metric, formulated as a second-order cone program.

  • PyScLERP: Python implementation of Screw Linear Interpolation (ScLERP) for manipulation planning, with examples such as pivoting and sliding.

  • Task-Oriented Grasping from Point Cloud Representation: neural network-based task-oriented grasp synthesis on object point clouds (IROS 2023).

Updates