“Hiring and retaining workers is one of the biggest challenges we face in running a warehouse,” says Michael Pultke, Head of Logistics at Obeta, the German electrical supplies wholesaler. The company's distribution centre is located at its headquarters in Ludwigsfelde, on the outskirts of Berlin, where it faces stiff competition for employees from numerous other warehousing and manufacturing facilities. This situation has been intensified by the COVID-19 pandemic, as Pultke explains: “We hire workers from nearby countries such as Poland and the Czech Republic but the borders have been closed intermittently, so it’s been even harder than usual.”
Obeta has over 65 locations across Germany serving 28,000 customers with a range of 850,000 items, of which around 28,000 are always in stock. Operating 24 hours a day for five days a week and shipping out more than 5,000 orders a day on a next-day delivery service, the lack of skilled staff has been a major headache. However, Obeta has a reputation for innovation – launching its online shop as early as 1995 and recently building a mobile app with a guaranteed delivery time of less than two hours within Berlin – so the company naturally turned to technology to solve its labour shortage.
Enter the latest generation of the Pick-it-Easy Robot, the brainchild of collaboration between Austrian logistics automation provider, KNAPP, and the Californian AI start-up, Covariant. The two companies met at an AI conference in Silicon Valley. "From the get-go, the chemistry was right both content-wise and on a personal level," says Peter Puchwein, Vice President of Innovation at KNAPP. "We decided to conduct some tests together to challenge Covariant’s artificial intelligence. The results were astonishing, so we decided to become partners and combine our expertise in robotics and logistics with the Covariant Brain." This phrase – the Covariant Brain – is the moniker the US company gives to its AI solution that allows robots to interact dynamically with the world around them. It has been developed using various types of AI including reinforcement learning, imitation learning and meta-learning. This intelligence means that Pick-it-Easy Robot excels in tasks such as 3D perception, few-shot learning and real-time motion planning, enablilng it to learn very rapidly how to handle unfamiliar items.
The Pick-it-Easy Robot system was installed at Obeta as part of an automated logistics solution from KNAPP that included the supplier's OSR Shuttle automated storage system. With Obeta’s warehouse receiving hundreds of new products each week – from light bulbs and electrical sockets to smoke detectors and circuit breakers – the company needed a system that could easily handle new items without disruption. Obeta was impressed that the Pick-it-Easy Robot learned quickly to pick products it had never seen before and was also excited by the prospect of this intelligence increasing exponentially through robots around the world sharing their learning with each other.
Fast and accurate
Obeta's robot currently works alongside its manual fulfilment stations, picking and placing products from storage containers into order boxes. The solution was deployed with minimal downtime and started picking orders within days. It achieves a peak speed of 600 items per hour, depending on order flow, which is equivalent to the manual workstations – although, of course, it can maintain this speed over a much longer period. The robot has been operating for just over a year, working up to 14 hours a day. It currently handles 70% of pickable SKUs and picks with an accuracy of 99%, despite having to learn new items each week. Michael Pultke says the robot has become an integral part of the company's warehouse operations: "We can depend on it to do its job, and don’t have to worry about it not being able to work, which is a huge win for us.”
With the COVID-19 pandemic having boosted e-commerce, where orders require more processing, there is now considerable demand for robotic fulfilment solutions to raise productivity, as well as to enhance accuracy to minimise returns. With solutions such as Pick-it-Easy Robot available in a Robotics-as-a-Service (RaaS) model, there is arguably more incentive than ever to invest in robotic picking.