It started, like most ambitious projects do, with a deadline we weren't ready for and a problem we didn't fully understand. ISRO's Robotics Challenge asked teams to build an autonomous drone that could navigate indoors — no GPS, no external beacons, just onboard sensors and whatever intelligence you could code into it.
We said yes before we knew what SLAM was.
The Problem
GPS works great outdoors. Satellites, triangulation, done. But take a drone inside a building — or into a cave, a warehouse, a disaster site — and you lose that signal entirely. The drone has to figure out where it is using only what it can see and sense.
This is the domain of Simultaneous Localization and Mapping, or SLAM. The drone builds a map of its environment while simultaneously tracking its own position within that map. It's a chicken-and-egg problem that's been studied for decades.
Where We Started
We had no hardware. No prior experience with ArduPilot. No depth cameras. What we did have was a team that was stubborn enough to spend weekends reading research papers and soldering components we'd ordered off Amazon.
The first month was pure research — understanding ROS, evaluating SLAM algorithms (ORB-SLAM3, RTAB-Map, LSD-SLAM), and figuring out which sensors we could actually afford. We settled on a RealSense D435i depth camera and a Pixhawk flight controller running ArduPilot.
What's Next
We're still building. The drone flies, the SLAM works (mostly), and the path planning is getting better every week. I'll write more about the technical details — the integration nightmares, the sensor fusion headaches, and the moments where everything just clicked.
If you're building something similar, or just curious, reach out. I'd love to talk.