The problem
Warehouse racking inventory is counted by people on ladders and scissor lifts. It is slow, it is periodic, and it is the kind of task where accuracy quietly degrades the longer someone does it.
Built for the Inter IIT Tech Meet 14.0 problem statement on autonomous warehouse inventory.
What it does
An autonomous mobile robot that drives warehouse aisles and scans racking from floor to top shelf.
- Jetson Orin Nano compute, running ROS2 with Nav2.
- RPLidar A1 SLAM with AMCL localization — the robot builds and holds its own map of the aisle.
- Four-channel ultrasonic dynamic obstacle avoidance, because a warehouse contains people and people move.
- Accuracy: ±7 cm horizontal, ±0.02 cm vertical.
The scanning mechanism
The part that made it work was mechanical, not computational.
- A belt-driven Z-axis scanning from 200 mm to 1800 mm, driven by a NEMA 23 — so a single robot covers a full rack face without needing to be tall.
- 100% QR code detection across the scan range, using Pyzbar.
- LoRa wireless emergency stop with under 200 ms latency. A machine that moves around people needs a stop that works when Wi-Fi does not.
- Hot-swappable 6S LiPo BMS, so the robot does not stop working to charge.
The vertical accuracy figure is the one worth reading twice: ±0.02 cm on the Z-axis is what makes QR detection reliable at 1800 mm, and it came from the belt drive and frame stiffness rather than from the software.
Where it went
Bronze Medal, Inter IIT Tech Meet 14.0, among all IITs. Hardware design lead.
Also 4th position, Engineers Conclave at the same meet, as presenting author on a paper on laser micro-3D printing of silicon carbide microstructures.