Attendees and industry members eye RoboBusiness 2025 as Silicon Valley prepares to host discussions on robotics advancements. Physical AI and simulation-driven learning are at the center of this year’s event, offering insights into rapidly evolving approaches that affect fields from e-commerce to healthcare logistics. Broadening beyond standard automation, the conference is expected to provide both technical and practical perspectives on how robots such as Ambi Robotics’ AmbiStack are reshaping warehouse operations. Organizers are also focusing on making sessions accessible for professionals aiming to solve practical issues in robotics deployment, while also addressing data and reliability concerns.
Earlier years highlighted growth in robotics logistics, but less emphasis was placed on combining simulation, reinforcement learning, and real data for training. Many prior exhibitions introduced automation technologies but rarely included substantial focus on large-scale AI foundation models or simulation-based adaptive learning. Compared to previous editions, this year’s event dedicates more resources to understanding “Sim2Real” techniques and foundation models in automation. Previous coverage of Ambi Robotics primarily showcased picking technology, not the complex manipulation and stacking capabilities that are now in focus with AmbiStack.
How Does AmbiStack Use Simulation?
AmbiStack, developed by Ambi Robotics, utilizes simulation-based reinforcement learning to adapt quickly to handling randomly assorted items, similar to solving a real-world 3D puzzle. Its AI engine is first trained in digital environments where the robot repeatedly stacks virtual objects, earning feedback and performance scores before interacting with physical goods. This strategy is designed to accelerate deployment in logistics and improve error correction as robots transition from simulated to real tasks.
What Expertise Drives Physical AI Debates?
Sessions at RoboBusiness featuring UC Berkeley’s Prof. Ken Goldberg and Ambi Robotics CTO Jeff Mahler will discuss research and product applications. Goldberg leads the AUTOLab and holds several institutional roles, while Mahler directs AmbiOS, the operating system enabling complex robotic manipulation. Both experts plan to address the scalability and data challenges of physical AI, providing live examples of AmbiStack’s decision-making and learning strategies.
“Physical AI blends data science, robotics, and real-world feedback for robust automation,”
Goldberg remarked on the forum’s agenda.
Who Attends RoboBusiness and What’s New?
RoboBusiness gathers engineers, developers, suppliers, and business professionals from all over the globe. In addition to the launch of the Physical AI Forum, the event features over 100 exhibitors, more than 60 speakers, as well as specialized networking and start-up activities. This year, registration is now open for events ranging from multi-modal AI panel sessions to direct demonstrations of industrial, e-commerce, and healthcare robotics.
“AmbiStack’s development proved the value of training in simulated environments before deploying new robots,”
explained Mahler, pointing to the focus on addressing reliability and speed.
With RoboBusiness 2025 bringing foundation models and Sim2Real techniques to the spotlight, industry professionals will have opportunities to scrutinize the efficiency and adaptability of AI-trained robots. Compared with earlier logistics robotics solutions, AmbiStack and PRIME-1 mark a transition toward more generalizable and scalable warehouse tools. Event organizers’ attention to technical sessions, data collection, and application-specific topics presents concrete learning takeaways for attendees. For businesses and developers alike, understanding the intersection of simulation, real-world robotics, and AI may be crucial to staying competitive in the logistics automation sector.