WorkCell

Problem

3D printers still need a human: swapping plates, clearing parts, queuing jobs, recalibrating. Running a multi-printer farm continuously and unattended means solving calibration, part handling, storage, and scheduling as one integrated system.

Approach

Build the whole cell as a closed-loop system. A gantry serves dual printers; computer vision with AprilTags and dynamic camera tracking keeps the gantry calibrated automatically; a centralized queue coordinates print execution, automated part removal, and shelf storage.

What I built

• Closed-loop auto-calibration system — a novel approach using computer vision + AprilTags with dynamic camera tracking to achieve 0.2 mm gantry positioning accuracy.
• Dual printers with a plate-swapping mechanism, storage automation, and shelf management.
• Continuous print queue system for unattended printing — centralized queue management integrating print execution, automated part removal, and storage.
• Inventory-optimization algorithms that automate shelf-placement logic based on print-height constraints and real-time storage availability.
• Tight coordination of firmware, sensor/actuator control, communication protocols, backend software, and mechanical integration.

Result

0.2 mm gantry positioning accuracy from the vision-based auto-calibration. Gold Award (Workcell category) and the Honda Innovation Award in Manufacturing at the National Robotics Challenge 2025, and selected to exhibit at OpenSauce 2026.

Media