PITTSBURGH — For years, robotics demos have been built around moments: a robot picks up an object, walks across a room or completes a carefully defined task.
At an IROS 2026 workshop in Pittsburgh on Sept. 27, researchers and engineers focused on a harder question: Can a robot keep working for eight hours?
Pimax took part in the full-day workshop, “Full-Shift Robot Co-Workers: Active Perception and Interaction for Human Environments,” which brought together more than a dozen researchers and industry experts working across robotics, computer vision, Physical AI and human-robot interaction.
IROS, short for the IEEE/RSJ International Conference on Intelligent Robots and Systems, is one of the world’s leading robotics conferences and is commonly grouped with ICRA and RSS among the field’s major international gatherings.
The eight-hour question reflects a broader shift in robotics. As machines move beyond controlled laboratory settings and into factories, warehouses and homes, the challenge is increasingly about what happens over time.
A robot may perform a task correctly once. Over the course of an entire shift, however, small errors can accumulate, people and objects move, and unexpected situations emerge. Staying useful for hours at a time requires a system to keep sensing, deciding and acting reliably even as those conditions change.
Those issues shaped discussions around four broad areas at the workshop: perception, reasoning, action and endurance. Participants included researchers from Carnegie Mellon University, Stanford University, the University of Toronto, Seoul National University and ETH Zurich, as well as experts from Tesla, Apple and Amazon Robotics.
Several talks also turned to a related challenge: teaching robots from human behavior.
Robots still have relatively limited access to high-quality data showing how people move through real environments, handle objects and respond to the world around them. As humanoid robots become more capable, that kind of data is becoming increasingly valuable for training systems meant to operate alongside people.
That is also where Pimax sees a connection to its own work.
The company has spent the past decade developing spatial-computing and near-eye display technologies and has expanded its global R&D presence through Pi Labs. Pimax has also joined the Hive Data Co-creation Initiative launched by Maniformer, and is exploring a range of first-person, multimodal data-capture devices, including head-mounted systems, grippers and data-capture glasses.
The goal is to capture more of what happens in real physical interactions: what people see, how they move and how they manipulate objects.
For Pimax, the move from spatial computing into Physical AI builds on many of the same underlying technologies. Systems that once helped people perceive and interact with digital 3D environments are increasingly being adapted to help machines interpret the physical world.
