Opinions expressed by Digital Journal contributors are their own.
Embodied intelligence — the idea that robots can learn and act through real-world interaction — was the defining theme at this year’s IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2025). Among the companies pushing the field forward, AgiBot drew major attention with the debut of its new humanoid robot, the AgiBot G2, and the finals of the AgiBot World Challenge 2025.
At its exhibition booth, AgiBot showcased a full lineup including the G2, OmniHand, Lingxi X2, and YuanZheng A2, giving attendees a close-up look at how embodied AI is moving from research to deployment.
The AgiBot G2, which launched earlier this month, was designed for industrial applications requiring high precision and human-like dexterity. Equipped with dual seven-degree-of-freedom arms, a three-DOF waist, and folding legs for coordinated motion, the robot demonstrated tasks such as force-controlled assembly, parcel handling, and a teleoperated archery routine with near-zero latency.
The company says that embodied intelligence is no longer just an academic term, and claims that they’re “showing how intelligent robots can adapt to complex environments and work safely alongside people.”
AgiBot also hosted the AgiBot World Challenge 2025, which drew 431 teams from 23 countries. The competition tested robots on real-world manipulation tasks, from folding T-shirts and operating kitchen appliances to handling moving conveyor belts, using AgiBot’s open-source dataset and EVAC baseline model.
Teams from the Chinese Academy of Sciences, HiDream.ai, and Tsinghua University & Shanghai AI Lab took top honors in the two tracks. Overall, all of the winners secured more than $100,000 in total prizes and innovation vouchers worth up to $100,000
This year’s IROS spotlighted a growing focus on integration, linking simulation, hardware, and AI into cohesive, reliable systems. AgiBot’s ecosystem approach, combining large-model intelligence with deployable hardware and open collaboration, illustrates how embodied AI is evolving from concept to industry reality.
