RUDRA, GPS-denied aerial autonomy
ISRO Robotics Challenge, URSC · Mar 2024 – Nov 2025
Final Field Round, IRoC-U 2025
Abstract
The task was to find a safe landing spot in a Martian-analogue arena and land on it with no human input and no GPS. Five of us built RUDRA as our entry to the ISRO Robotics Challenge. I built the drone and its perception stack. It reached the Final Field Round and placed in the top sixteen teams in India, and it failed there in a way that set my research direction.
The system
I coded the ORB-SLAM stack for localization, and wrote the algorithm that used LiDAR and Intel RealSense depth data to pick out safe landing spots and arena boundaries. There was no terrain to test on, so I designed and built the arena we flew in.
What happened at the finals
At the finals the arena was enclosed by a safety net. At our tested speed the drone struck it. We lowered the speed and it cleared the net, but the battery depleted before it could land on a safe spot.
A drone that cannot absorb a net in a controlled arena cannot survive Mars, and that showed me how much deeper the problem ran than I had thought. The mission had been planned around a single operating point, with no way to replan when the arena was not what we had assumed.
Those nine months were my first real taste of research, and the failure is the reason I now work on perception-driven decision making under uncertainty.