Brightpick unveils details about machine vision and AI behind its autopickers

Brightpick, a provider of warehouse automation solutions for order fulfillment, has unveiled new details about the proprietary machine vision and AI that are “critical for the seamless operation of its Brightpick Autopicker robots”.

The technologies are used for everything from scanning totes (bins) to create 3D models for future picking of items, to determining ideal picking angles for both totes and individual items.

The company also elaborated on its Brightpick Intuition software and how it keeps the entire robot fleet precisely orchestrated and optimized for maximum throughput in ecommerce and e-grocery warehouses.

All of the advanced technologies work together to ensure Brightpick Autopickers deliver strong throughput for its customers.

Jan Zizka, CEO and co-founder of Brightpick, says: “Technically, the software is the most highly engineered and valuable part of Brightpick Autopicker.

“Using our world-class machine vision, AI and fleet optimization software, our talented engineers created a very precise and powerful robotic system for order fulfillment in warehouses.”

Brightpick Autopicker is an autonomous mobile robot that moves around the warehouse, retrieves product storage totes from shelving, and robotically picks items from those totes to consolidate orders directly in the aisles.

The versatile robot can also be used for tasks such as Goods-to-Person picking, pallet picking, stock replenishment, dynamic slotting, order consolidation, buffering and dispatch.

The patented Brightpick Autopicker can pick a diverse range of products using its proprietary world-class machine vision and advanced AI algorithms which have been trained on more than 500 million picks to date.

The machine vision, originally developed by Brightpick sister company Photoneo, is used to scan totes and create virtual 3D models for subsequent picking of items. AI is then used to determine the most optimal item to pick and the exact pick point where to pick the item.

The robotic arm on Brightpick Autopicker picks items using suction cups, which incorporate sensors that measure how much pressure is needed (depending on the weight of the item).

Brightpick Autopicker uses a two-axis SCARA robotic arm, which costs only 10% of what a six-axis robotic arm would cost but performs just as well due to the combination of machine vision, AI and versatile suction cups.

AI also enables the Brightpick Autopicker robots to generalize related data to pick previously unseen items without additional training. The robots can reliably pick ambient and chilled groceries, pharmaceuticals, medical devices, packaged goods, cosmetics, electronics, polybagged apparel and more.

In addition, machine vision in Brightpick Autopicker can be used to gently place any picked item (from a height of 1 mm) into the order tote, which is useful for fragile items. It can also make sure that an order tote is not overflowing and place newly-picked items in an appropriate empty spot.

Brightpick Autopicker also uses AI and machine vision to identify the optimal gripping points to retrieve the totes themselves.

This is useful in warehouses that have uneven floors, even by an inch, because it enables the robots to seamlessly adjust their position to grip totes at the optimal spot – even when the totes are not perfectly aligned on the shelves due to uneven floors.

This is relevant for existing warehouses because it enables Brightpick to easily install its robots without any modifications to flooring.

All fleet management and fulfillment workflows are handled by Brightpick Intuition software. Brightpick Intuition uses “time-space planning” to optimize the speed and throughput of the entire fleet of robots.

To maximize overall efficiency, it plans the paths of the Brightpick Autopicker robots in a way that ensures they do not cross each other, which would impede or slow them down.

If a robot must cross the path of another robot, the software ensures one robot “gives way” to allow the higher-priority robot to pass.

Moreover, if a Brightpick Autopicker robot is blocking an aisle because it’s picking there, other robots will go around it to avoid being stuck in that aisle (even if that aisle is the shortest pick path for the other robots).

On average it takes only a few seconds to re-plan the entire robot fleet if something unexpected disrupts the precisely planned workflows (for example, if a human needs to walk into a robot’s path).

Each robot is also able to act autonomously in critical situations, for example quickly activating collision avoidance when a human walks in front of them.

Brightpick Intuition also enables Brightpick Autopicker robots to dynamically slot (place) each item in the warehouse in the most optimal storage location based on each item’s historical or expected frequency of being picked (ordered by customers).

The robots do the slotting mainly at night, but also throughout the day during any quiet periods for the robots.

Brightpick Intuition orchestrates and optimizes all fulfillment workflows in the warehouse, including picking, consolidation and replenishment. The software easily integrates with a warehouse’s WMS or ERP system using standard APIs.