Plus One Robotics a provider of parcel handling robotics platform has published a video of their PickOne vision system in operation at Pitney Bowes’ BWI e-commerce fulfillment hub. Developed collaboratively by Plus One Robotics and industry frontrunners. PickOne provides robotic arms to work in a range of pick-and-place tasks within e-commerce fulfillment and distribution hubs.
The Operational Workflow of PickOne
Step 1: Image Capturing and Analysis
Upon the arrival of items in the picking area, the PickOne Perception Kit captures detailed images of the items.
Step 2: Identification
Leveraging 2D, 3D, and AI algorithms, PickOne identifies each pickable item within the scene, assigning a corresponding confidence score to each identification.
Step 3: Communication
Utilizing the PickOne API, the software transmits an array of pick locations, poses, dimensions, and distinctive data for each pickable item to the robot controller. In instances where items lack sufficient confidence scores, PickOne triggers a Yonder (a remote supervisory software platform) request, swiftly connecting with a remote Crew Chief for prompt exception handling.
Step 4: Execution and Learning
Simultaneously, Yonder preserves the Crew Chief’s response, enabling machine-learning algorithms to continually enhance system intelligence. As the process continues, PickOne sends the collected data to the robot, ensuring optimal execution. This iterative learning loop contributes to elevated performance levels over time.
Software Platform Specifications
- Typical Response Time from Pick Request to Response: 250ms – 480ms
- Supported Sensors: Intel RealSense D415, Intel RealSense L515, Zivid One
- Max Number of Sensors: 2
- Communications Methods: SMS, Fanuc Enhanced Vision Interface
- Supported Robot Controllers: Fanuc, Yaskawa, ABB,
Universal Robot, Kuka, Kawasaki, Denso, Festo,
Rockwell Automation (Allen-Bradley)