This repository provides a modular post-processing framework developed by the Virginia Commonwealth University Additive Manufacturing (VCU AAM) research group. It integrates computer vision, machine learning, and robotic path planning to automate powder cleanup of laser powder bed fusion additive manufacturing machines.
- Intel RealSense D435i
- Universal Robots UR5e
| Module | Description |
|---|---|
| am_vision | RGB and depth frame capture utilities. |
| ml_vision | Machine-learning interface layer that performs model inference (e.g., YOLOv5) to locate and distinguish build and feed cylinders. |
| path_planner | Grid-based motion planner that generates coverage paths around detected obstacles in a rastering pattern. |
| robot | Interface layer that executes robot motion commands, communicates with the robot, and collects robot position data. |
Each module can operate independently, but they are designed to form a cohesive vision-to-motion workflow.
am-powder-cleanup-workflow/
│── am_vision/ # Image alignment and preprocessing scripts (to be refactored)
│── ml_vision/ # ML-based object detection and mask generation
│── path_planner/ # Grid-based path planning and offset handling
│── robot_control/ # Robot execution and motion interface
│── contributing.md # Contibution guidelines
│── master_script.py # Example entry point
│── README.md # This overview
- Python 3.10.X
opencvnumpymatplotlibur_rtde
VCU Additive Manufacturing Research Group
[email protected]