Pallets are brought to the robotic depalletizer/Pick Station by either human workers or a pallet conveyor. In fully automated environments, the pallet is scanned, and product information sent to the warehouse management system to keep track of inventory.

The PickOne Perception Kit images the pallet. The software then analyzes the 2-D, 3-D, and AI data to identify each pickable item in the pallet and assigns each one an associated confidence level.

PickOne sends the robot controller an array of pick locations and poses for each pickable item via the PickOne API.

For items with low confidence levels the software automatically generates a Human-in-the-loop(HITL) request by calling in remote human intervention. With a response time of under six seconds, these robot managers efficiently handle exceptions such as reflective packages or unfamiliar shapes by simply selecting an item in the scene to be picked. In seconds, our Human-in-the-loop (HITL) updates PickOne, and PickOne sends the data to the robot.

In parallel, Human-in-the-loop (HITL) stores the robot manager’s responses, enabling the machine-learning algorithms to make the system smarter as it works. This ensures continuously increasing performance over time.

Items are transferred onto an outfeed conveyor,sorter, or specialized equipment like a descrambler that separates and positions them for subsequent processes like storage, further processing, or fulfilling outbound orders.