DexRobot Showcases Advanced Robot Hands and AI Development Platform at IROS 2026

Pittsburgh, PA – DexRobot, a full-stack embodied AI company, is showcasing its advanced robotic manipulation capabilities at IROS 2026 (Booth 507). The company is presenting three accepted research papers, its flagship DexHand021 Pro with full-palm multimodal sensing, and the DexGEM development platform.

The DexGEM platform represents a significant research milestone in physical AI dexterity. It is the basis for one of DexRobot's accepted research papers focusing on transforming human demonstrations into contact-rich robot policies. The platform learns dexterity from human hand movements captured via motion capture and first-person video. These demonstrations are reconstructed in simulation, adapted for robotic hands, and then transferred to real robots without teleoperation or specific robot training data. The development includes tasks involving two-hand coordination and continuous contact manipulation.

The company's flagship product, the DexHand021 Pro, offers research-grade dexterity for real-world tasks. It features an integrated wrist and hand architecture with 22 degrees of freedom, enabling natural wrist-hand coordination. Its multimodal sensing system covers the palm, fingertips, and finger sides, including a high-resolution tactile array and a wide-angle camera. A dedicated edge AI chip closes the perception-compute-control loop, allowing for millisecond-level decision-making.

DexRobot integrates hardware, perception, data, models, and scenarios into a cohesive pipeline. The company aims to create a complete portfolio of robotic hands capable of converting human demonstrations into training data and skills. By validating each step in real-world scenarios, the company closes the development loop from hardware to physical tasks, as demonstrated by the DexGEM platform on actual robots.