Nvidia’s GTC event is now the “go-to” AI show. While the event is filled with technical sessions that talk about the latest and greatest tech, it also has several industry sessions that demonstrate how the technology is being used.
One of the more interesting retail breakout sessions was the session, “Leverage Physical AI to Simulate and Orchestrate Robotic Fleets for Retail Fulfillment Centers,” featuring Germany-based Otto Group.
The retailer has embraced physical artificial intelligence and is moving it into real-world operations. However, that shift is showing up as a coordination challenge. The international retail and services company envisions a future in which robots work together in highly automated warehouses, and people oversee the process rather than performing repetitive tasks.
The complexity of a global supply chain
Otto Group serves nearly 45 million customers worldwide and generates around €15 billion in revenue.
One of the company’s defining characteristics is that it owns and manages its entire supply chain. Running the supply chain end-to-end gives Otto control over what customers see, but it also entails greater complexity behind the scenes. That’s why much of the work is happening in supply chain and logistics, especially around AI.
“This is a very crucial moment. A lot of things had to be done in the last few years, working on incrementally optimizing processes and warehouses. Right now, it seems to be a moment where we can make very big shifts,” said Martin Umland, vice president of supply chain management at Otto Group.
From optimization to ‘big shifts’
Umland shared more details about what that shift looks like during his session. He said Otto Group’s warehouses already run a mix of robotic systems, including autonomous mobile robots (AMRs) that move goods, shuttle systems, and robotic arms that handle specific tasks.
These systems were added over time from different vendors and for different use cases. They serve their purpose, but they don’t really work together. Instead of replacing them, Otto Group created what it calls a coordinated autonomy layer (CAL), which brings together different robotic systems so they can operate as part of a single workflow.
Mapping the digital twin
The project started with mapping a pilot warehouse.
Otto Group used Boston Dynamics’ Spot robot dog to move through the aisles and scan the space, collecting data on the entire environment. After about a week, there was enough data to create a 3D digital twin of the warehouse in Nvidia’s Omniverse computing platform.
The digital twin turned out to be more precise than Otto Group’s own documentation. It also changed how the company worked with the space. Instead of treating the warehouse as fixed, Otto Group could test different setups and adjust how things move through the environment before making changes on the floor.
Testing reality in simulation
From there, Otto Group moved into simulation. It ran comparisons between different layouts and tracked how robots moved through each scenario.
One area of focus was how often robots had to stop and restart their tasks. In one case, a revised layout reduced those stop-and-go events by about 20 percent, according to Stefan Borsutzky, co-CEO of Otto Group one.O GmbH, the company’s technology and consulting subsidiary.
“The material flow was optimized, and the productivity of the robots increased. So, we used this result and implemented the new handover area already in the warehouse,” said Borsutzky. “We are working towards a self-controlled warehouse under human supervision.”
The complexity is less about any single robot and more about how they interact. If multiple machines move in the same space in a warehouse, typically, one doesn’t adjust to what the other is doing. Bringing the robots into a shared environment is the goal.
That’s where CAL comes in. This layer sits between the warehouse management system (WMS) and the machines themselves. As Borsutzky explained, CAL “takes the request coming from the WMS” and assigns tasks across robot fleets. It also manages movement across the floor by routing robots and reducing the chances of incidents or collisions.
For now, CAL handles task assignment and routing. The next step is to add what Borsutzky called an “artificial brain,” an AI layer that takes in data and adjusts how tasks are assigned and how robots move. The digital twin in Omniverse would provide a way to see and validate what the AI is doing.
Otto Group is working closely with Nvidia, as well as other partners such as Reply and Google.
This “learning journey” continues for Otto Group, said Umland. The plan is to expand beyond individual warehouses across the company’s broader network in Europe. Internal teams are working through operational challenges alongside partners who provide infrastructure and technical support.
“We don’t know every single step in the next weeks or years, but we are sure that we have the right partners on our side,” Umland concluded.
He added: “We always talk about an island beyond the fog. Envision a very big warehouse where we are shipping millions of items per day, and where we have a lot of tasks that are done between unloading, loading, sorting, and bringing goods to other places. We are working on our vision. We are sure that robotics and AI are key success factors.”
Final thoughts
The Otto Group’s transition from manual labor to a Coordinated Autonomy Layer (CAL) represents a transitional moment in the evolution of logistics. By moving beyond individual “islands of automation” — where robots from different vendors operate in silos — the company is creating a unified digital nervous system for its warehouses.
This case study highlights three critical shifts in the industry:
- The Power of the digital twin: Using tools like Nvidia’s Omniverse to create models more precise than original blueprints proves that “guessing” on the warehouse floor is a thing of the past. Simulation enables risk-free experimentation, as evidenced by the 20% reduction in robotic stop-and-go events.
- A new role for human talent: As robots take over repetitive tasks, the human workforce isn’t being eliminated; it’s being elevated. The shift toward “oversight” roles suggests a future where warehouse staff become orchestrators of technology rather than manual movers of goods.
- Infrastructure as software: By treating the physical warehouse as a flexible environment that can be “reprogrammed” via simulation, Otto Group is proving that the most valuable asset in modern logistics isn’t just the hardware—it’s the data and the coordination layer that makes the hardware work together.
As physical AI continues to mature, the success of the CAL project will likely serve as a benchmark for other global retailers looking to solve the “coordination challenge” in an increasingly complex global market.
For more on how automation is reshaping the workforce, check out this report on Amazon’s latest robotics-driven layoffs.
