Research Project

GRASP: General Robotic Arm for Smart Packaging

Research Team

Lead Researchers:

Collaborators:

Student Team:

Funding

Background: Many processes in manufacturing and packaging rely on manual labor, leading to reduced effective working time and poor ergonomics. This motivates the need for automation to improve consistency, efficiency, and operator comfort.
Current System: MiniMed develops diabetes management solutions, producing 16,000 sensors per day. Packaging is performed manually across three shifts, each spanning six hours of effective working time. Each shift requires three workers responsible for box preparation, sensor insertion, and sealing.
Challenges:
- High production throughput requirements
- Real-time perception and control for 3-second cycle time
- Robust performance under variable object placement

Control: Develop a robust actuation control system to reliably execute a sensor packaging process, ensuring consistency with zero-failure tolerance over repeated cycles.
Perception: Design and implement a computer vision pipeline to detect, localize, and estimate the pose of sensors from camera input, extracting accurate position and orientation for robotic grasping and insertion operations.
Integrate vision and control systems into a unified workflow and develop comprehensive documentation to support deployment, operator training, and system handoff at MiniMed.

The overall approach was structured into three modules: Perception, Control, and Design.
Perception module:
- Environment detection using ArUco markers [1]
- MiniMed sensor detection using learning-based perception via YOLO [2] and FoundationPose [3]
Control module:
- Lightweight PyBullet simulation environment [4]
- Open-source Python SDK for control implementation
Design:
- MiniMed sensor arrangement strategy for grasping
- Customizing end-effector for multi-sensor grasping

3D Models: Current efforts have produced 3D scans of MiniMed sensors (shown on right) for control simulation and model-based perception.
Grasp Configuration: Empirical studies have identified optimal grasp configurations and hardware for the MiniMed sensors.
Integration of perception and control modules is under progress.

Application: Manufacturing and packaging automation.
Value: General robotic solution for human-safe deployment.
Potential customer: Successful deployment can lead to further automation opportunities with MiniMed.

Start Date:
End Date: TBD

Research Team

Lead Researchers:

Collaborators:

Student Team:

Funding