Home / News
Build log
Updates from the workshop: what we've built, what we've learned and what's coming next.
Product
What started as a robot concept with early image recognition has grown into BARCONA AI, a software core that turns warehouse images and video into inventory data.
With the core in place, we're introducing a second way to deploy it. The Detection Unit mounts a camera and compute module on existing forklifts, so inventory is captured during normal operations. A first test setup is in preparation.
Engineering
We've trained a recognition model on images from our own test setups instead of relying on generic models. It's noticeably better at finding barcodes on real warehouse labels, even at an angle or in poor light.
Engineering
We built a simulated warehouse hall to test how the robot drives, where it stops and how it scans each rack. Changes can now be tried in minutes before they go anywhere near hardware.
You can watch a simplified version of a scan mission on our homepage.
Engineering
At the beginning of April we started testing the full camera chain end to end against our new datasets: finding a barcode, moving the camera onto it and reading it reliably.
Later in the month we added a tilt routine: the camera stays on the robot, angles up the rack level by level and only keeps the sharpest frames. That's the scan pattern the TraceUnit will use in the aisle.
Engineering
We've started building our own datasets from real warehouse labels: different printers, angles, distances and lighting. They're what we test and train against from now on.
The camera also learned to pan, tilt and zoom on its own towards what it detects.
Engineering
The core architecture is in place: the foundation that takes in data from the camera, barcode reading and LiDAR and turns it into one consistent picture of the warehouse.
A first camera setup already picks out barcodes in live images. Next up: data, and lots of it.
Design
The idea has become a concrete plan: the TraceUnit, an autonomous robot that drives the aisles, stops at each rack and reads every barcode with its camera, while LiDAR keeps track of where it is.
Design
Stocktakes are slow, manual and out of date the day they finish. That's where Barcona started: with the question of whether a camera could do the counting instead.
Stay in the loop
Email us and we'll add you to our update list. We only write when there's real progress to share.
Subscribe by email →