Daimon Robotics launches RobOmni, the first omni-modal evaluation benchmark including tactile sensing for physical interaction in robotic manipulation

Daimon Robotics, an embodied intelligence company from Hong Kong, unveiled RobOmni at ICRA2026 as a standardized benchmark for evaluating tactile perception and dexterous manipulation. The platform is built on NVIDIA Isaac Sim and provides omni-modal information including high-resolution fingertip tactile sensing, wrist-mounted RGB vision, gripper status, TCP trajectories, action commands, and external camera observations.

RobOmni supports tactile ablation testing, allowing developers to evaluate robot policies with and without tactile information to measure how tactile sensing directly contributes to task performance. The benchmark evaluates multiple dimensions: task success rate, manipulation efficiency, dexterous manipulation capability, operation failure events (slip, jamming, collision, retry), and generalization robustness across dozens of manipulation scenarios such as grasping, placement, precision insertion, and component assembly.

The industry previously lacked a unified evaluation framework for tactile intelligence in contact-rich manipulation, making it difficult to compare systems or measure progress. RobOmni addresses this gap by providing a common standard and helping establish a data flywheel of data accumulation, capability validation, and model iteration to accelerate industry development.