I'm Rahul Rustagi. I earned my Masters in ECE from Georgia Tech, with a focus in robotics. My work centers on controls, embodied AI, and reinforcement learning—combining them to make AI systems more capable of reasoning and acting safely in the physical world. I am advised by Prof. Glen Chou.
I have spent some excellent summers as a research intern at Symbotic in Boston, Helicopter Lab at IIT Kanpur and Carleton University in Canada. I have published papers in top-tier venues such as AIAA and TCAS-II. My prior research has been covered by media outlets such as Hello-Robot Newsletter.
I am fortunate to be co-advised by Prof. Sonia Chernova during my Masters, Prof. Abhishek during my undergrad. I completed my bachelor's degree in aerospace engineering at IIT Kanpur, where I also earned minors in computer science and literature. Outside of research, I enjoy playing guitar, building robots, and hiking.
Autonomous landing of an aircraft on a ship deck, perturbed by the winds and sea waves, is an inherently difficult task. To achieve this, we not only need to touchdown on a very small area, but we also need to time it accurately for a safe landing.
We propose a deep reinforcement learning framework for MET scheduling in IoMT networks to mitigate energy holes and improve network lifetime and stability.
This work introduces an AoC-based formulation and a DRL scheduler for mobile energy transmitters in asynchronous IoT networks, improving charging efficiency over baseline methods.
A semantic mapping framework built on Khronos and ORB-SLAM2 for creating a semantic map using the Hello Robot Stretch mobile manipulator. This framework creates hierarchical object-centric maps, tracking moving objects, and maintaining long-term object changes providing a contextual understanding for household navigation and mobile manipulation.
news | codeAn end-to-end robotics pipeline that translates natural-language instructions into precise pick-and-place actions using a VLM-based planner.
github | demoA novel approach to compressing 3D Gaussian Splatting models by leveraging scene structure while preserving high-quality reconstruction.
paper | codeA generalized approach for 2D stereo matching using PDE-based optimization for robust disparity estimation.
paper | codeA centralized control framework for multi-agent systems to achieve pattern formation using a virtual leader-follower assignment.
problem statement | codeA medium-sized UAV that autonomously navigates through a field environment to pick up and deliver objects using object detection, path planning, and an electromagnetic gripper.
problem statement | github