After a successful pilot in Prague, Brightpick’s AI-powered warehouse robots will automate order picking, consolidation, and dispatch for Rohlik Group, one of Europe’s largest e-grocers.
The human hand is said to have 27 degrees of freedom — and as any robot designer will tell you, that’s not an easy thing to replicate cost-effectively.
“Replacing the human hand is extremely difficult,” said Jan Zizka, CEO of Brightpick. “I think some companies tried, and it ended up costing €50,000. You would need to sell billions of those hands to get to a reasonable price.”
Human evolution has granted us attributes that enable survival and advance our capabilities, such as picking and holding objects of irregular shapes and various textures. Though robots and automation technology have yet to catch up, e-grocery warehouses can usually get away with simpler motion — with vacuum grippers, for example — but not without challenges.
“In e-grocery, there’s a well-defined portfolio of items, but they’re still quite diverse. There are bags and boxes of different thicknesses. Some materials are porous and don’t work well or at all with a suction cap,” said Zizka. “Throughout our company’s history, we’ve probably designed around 200 grippers, not only for grocery, but also for very specific items in automotive, and some for as general of use as possible.”
Brightpick recently completed a nine-month pilot of its robotic solution in a Prague fulfillment center for Rohlik Group, one of Europe’s largest e-grocers with more than 1.5 million customers and 12 million orders per year. The solution includes autonomous mobile robots (AMRs) that fully automate order picking, consolidation, and dispatch. Rohlik plans to roll out the Brightpick solution across multiple warehouses in 2023 and 2024, starting with Munich in July 2023 and continuing with Frankfurt, Vienna, Prague, and other fulfillment centers. Once fully deployed in the Munich warehouse, the center will be able to dispatch more than 1,000 order totes per hour and 5,000 completed orders per day, where some orders contain multiple totes.
“Brightpick Autopicker is our flagship product with a relatively simple arm, which makes it a cost-effective solution,” said Andrey Bakholdin, VP of marketing and business development at Brightpick. “Even though the arm only has two degrees of freedom, the gripper, or the suction cap, that we designed can grab items at a 60° angle. With our AI and 3D-vision cameras, we can pick most ambient groceries and now chilled groceries.”
Item picking is a major operational challenge, as grocery orders contain an average of 30 items per order. A typical grocery fulfillment center also has various picking zones for ambient, fresh, chilled, and frozen items. Brightpick’s mobile robot solution automates both picking and order consolidation across these different zones. The robots are patented and powered by proprietary machine vision, AI algorithms trained on more than 250 million picks, and machine learning to improve with each pick.
“We use 3D vision and artificial intelligence to learn how to create better grippers. Our machine learning has seen enough items to generalize unseen items and make the robots smart enough that they don’t need to be retrained. This is the magic and probably the biggest milestone we’ve had thus far,” said Zizka.
Brightpick Autopicker robots are like humans with carts that move around the warehouse, retrieve product storage totes from shelving, and robotically pick items from those totes to consolidate orders directly in the aisles. The robots don’t need to travel back and forth to centralized picking stations. This results in faster picking, and they reliably pick most groceries with 99.9% accuracy. They also include goods-to-person capabilities for items that require human dexterity and autonomously find the nearest human picking station.
In most online grocery fulfillment centers, including Rohlik’s, picked items are sent via totes on conveyor belts to a central location for consolidation. Brightpick Dispatcher robots consolidate items arriving from various picking zones into completed orders to dispatch for delivery, curbside pickup, or in-store pickup. These robots can also pre-pick orders and store them in a temporary staging area until they need to be dispatched. This enables warehouses to smooth out picking volumes throughout the day and increase total daily throughput.
Brightpick Intuition software is the AI-powered brain of the solution that guides the robots from picking all the way to dispatch. A typical warehouse will have 15 to 100 robots, which companies can remotely monitor and control, tracking every robot, order, tote, and item in real time.
The fully autonomous, end-to-end robotic solution reduces picking labor by 95% and cuts costs for order fulfillment by half. However, humans are still needed in any warehouse regardless of how many robots there are and how much programmable intelligence they have.
“The human element is a huge differentiator for us because we provide an end-to-end solution,” said Bakholdin. “We don’t just do robotic picking — we also do storage, retrieval, consolidation, and so forth. And because we can always go to a human, as a last resort, we always get the order picked. There’s no single point of failure.”
Oftentimes, when providers install robotic pickers as a point solution and can’t pick certain items, humans must physically go to the robots to fix them, move the items, or restart the system. In this case, picking issues aren’t handled as part of the process flow and are considered failures.
Contrarily, with so much talk about collaborative robots and humans working efficiently with machines, Brightpick actualizes this concept by designing a robot’s inabilities into the workflow. Humans are included in fallback scenarios that keep operations running, so much so that there don’t appear to be any issues on the floor. And the machine-learning algorithms can collect information to make the robots smarter for next time.
The icing on the cake — and what e-grocers such as Rohlik love to hear — is that the solution takes mere weeks to deploy.
“There are three reasons we can deploy so quickly,” said Bakholdin. “First, we use standard commodity shelving and totes that can be procured anywhere, so they are very easy to assemble. Second, our robots use LiDAR for navigation, which is purely visual, so you don’t need any magnetic tape, QR codes, or fencing around the system. And the third part is probably the most important: All the heavy lifting — designing the solution, simulating it, and ensuring that it works — happens before the installation begins.”
The company uses a warehouse’s real-life historical data to create a detailed simulation and test the solution extensively to visualize the system’s behavior before installation. The team designs a digital twin of the facility and robot fleet and delivers an animation of the process and a detailed report with metrics. This simulation saves upfront time and costs to ensure the solution meets physical requirements and performance expectations.
“All the robots are standalone, self-sufficient, and equal,” said Zizka. “The entire idea is that there’s no single point of failure. If someone turned off or removed all the robots in the warehouse for some reason, the system would still work. Performance would be lower, but it would work exactly as before. We’d bring the robots, connect them to the Wi-Fi, and they’d run. It’s like plugging a USB stick into a laptop.”
Brightpick Autopickers were designed for e-commerce and grocery retailers and are suitable for large, small, and micro-fulfillment centers. They work with standard warehouse shelving and totes, enabling fast deployment and easy integration with any warehouse environment, including existing operations and mezzanines.
You may also like:
Filed Under: Warehouse automation, ALL INDUSTRY NEWS • PROFILES • COMMENTARIES, Robotics • robotic grippers • end effectors