Blog | 06 November, 2024

Transforming Logistics: The AI Warehouse Robot Simulator Experience

The AMC Bridge Research Department team has introduced a new technology demonstration: AI Warehouse Robot Simulator. This demo showcases the simulation, suggestion, and training of AI-driven robots tailored to warehouse technical requirements. By utilizing warehouse models, the simulator aims to enhance efficiency, accuracy, and adaptability in logistics, supply chain operations, and warehouse management while also boosting workforce safety and minimizing the risks of workplace injuries.

Projects like these serve educational and marketing purposes within the company, offering colleagues the chance to expand their skills and gain valuable experience. This time, our team explored a range of NVIDIA® Omniverse capabilities. Oleksandr Myroshnychenko, Project Manager in the Training and Research Department, shared insights into the challenges and opportunities Omniverse presents in this innovative project.

Combining Autodesk Revit and NVIDIA Omniverse: A New Solution

This tool was developed by the Research Department, making it something entirely new, and our first goal was to prove the concept. So, what problem are we solving? We're combining two tools. On one side, there's Autodesk® Revit®, a well-known BIM suite where engineers design everything from buildings to skyscrapers, stadiums, airports, and railway stations—essentially anything related to architecture. On the other, we have NVIDIA Omniverse. NVIDIA, the world's largest producer of graphics cards, created Omniverse as a comprehensive platform for their GPUs. It's packed with features for neural network training, photorealistic rendering, simulations, robotics, and much more. Our project brings Autodesk Revit and Omniverse together through a plugin that exports data from Revit to Omniverse, enabling further work there.

Simplifying Robotics for Engineers

In Revit, an engineer might design a large warehouse where robots, instead of people, will operate. The engineer sets up the layout, positions shelves, and uses our add-on to export the 3D model of this warehouse to Omniverse. There, we use a feature called Isaac Sim™, which is designed specifically for robot programming. In Omniverse, we can set up robots to operate within this space, defining loading and unloading points and basically letting them carry out their tasks autonomously.

The unique aspect of this demo is that we've managed to simplify the entire process. The Revit engineer doesn't need deep knowledge of robotics—with just a few clicks, they can set up and run the necessary scenarios.

AI Warehouse Robot Simulator Features:

  • Exporting Autodesk Revit warehouse model geometry for training and simulating AI-based robots in NVIDIA Isaac Sim.
  • Training AI-based robots according to warehouse constraints.
  • Automatic simulation of AI-based robot operations.
  • Obstacle-avoidance navigation for robots.
  • Interactive testing of trained robots.
Reducing Costs and Enhancing Safety

This solution enables efficient robot programming for closed spaces with specific tasks, making it safer and more cost-effective. With no humans in the warehouse, there's no risk of injury, and robots take care of all tasks. Additionally, it significantly reduces costs. Traditionally, different teams handle building design, robotics, and warehouse organization. Now, all these complex tasks are automated—the Revit engineer designs the space then organizes the rest with just a few clicks.

Learning Curve: Adapting to Omniverse's Capabilities

Working on this project was both challenging and rewarding for our team, though it certainly demanded advanced knowledge and a high learning curve. The project has two main parts: working with Revit and working with Omniverse. We have extensive Revit experience within the company, so our team was already familiar with Revit's modeling and API for exporting data, geometry manipulation, and data translation to Omniverse.

Omniverse presented some of the project's biggest challenges, requiring our team to examine its capabilities thoroughly. And it's not just about programming. When programming a robot's movement, it's essential to understand basic physics and mechanics. The robots we selected for our project turn by adjusting the speed of their wheels rather than simply steering in a particular direction. This requires us to calculate movement dynamics to coordinate wheel speeds accordingly. Of course, this isn't always the case—many other robots turn more like cars, which makes the process simpler to implement. Each robot has mass and inertia, so we need to account for these to prevent it from speeding up uncontrollably and crashing into objects.

In addition, we've implemented rules within the warehouse environment, essentially creating a set of "traffic rules" for the robots. They're programmed to avoid each other, navigate around shelves and pillars, and reach loading or unloading points from precise angles. Although we didn't fully use neural networks, we applied them to help robots detect potential collisions, adjust paths, and avoid each other in real-time. Each of these factors required a careful application of physics and mechanics principles to ensure smooth and safe interactions within the warehouse.

Omniverse also enables photorealistic rendering, allowing us to incorporate visual enhancements that breathe life into Revit's basic gray warehouse model. In Omniverse, the warehouse now features realistic reflections on the floor, dynamic shadows from moving robots, and richer textures that create a more immersive experience. While it may not yet fully replicate the real world, this transformation marks a significant leap from the basic Revit model, showcasing the potential of our technology to simulate real-world conditions.

Our team was genuinely inspired while working on this project. I thrive on initiatives that challenge us to think outside the box and allow for the acquisition of complementary skills, and this project offered numerous opportunities for just that. Additionally, we are actively exploring the potential of AI for key industries served by the company, aiming to harness its capabilities to the fullest.

All third-party trademarks belong to their respective owners. For more details, please refer to the Third-Party Trademarks list on the Privacy and legal notices page.

Return to blog page

Subscribe to our news

We will keep you updated with the latest news

scroll down to explore
to the top