DexRobot Unveils Advanced DexHand021 Pro at IROS 2026

PITTSBURGH, Sept. 30, 2026 /CourierPR/ -- DexRobot, a full-stack embodied AI company focused on dexterous manipulation, is showcasing its cutting-edge capabilities at the International Conference on Intelligent Robots and Systems (IROS) 2026, held in Booth 507. The company is presenting three accepted research papers, along with its flagship DexHand021 Pro, a highly advanced dexterous manipulation system.

The DexHand021 Pro is a research-grade dexterous hand designed for real-world tasks. It features 22 degrees of freedom (DoF) in a wrist-hand integrated architecture, enabling natural wrist-hand coordination for tasks such as turning, aligning, flipping, and tool use. The hand's dual-tendon bidirectional drive with active distal joints allows for independent active flexion and extension, providing precise force control and passive compliance.

At the heart of the DexHand021 Pro is a full-domain multimodal perception system. The hand's tactile envelope array offers 0.01 Newton (N) normal-force resolution at 200 Hz sampling, allowing for highly sensitive tactile feedback. Additionally, a 5-megapixel, 160-degree wide-angle palm camera tracks contact, force, and slip, guiding adjustments and enhancing the hand's dexterity. These advanced features are supported by a dedicated edge AI chip that closes the perception-compute-control loop, enabling millisecond-level decisions.

DexRobot's DexGEM platform is a research milestone in physical AI dexterity. It learns dexterity from human hands by reconstructing demonstrations captured through motion capture and first-person video in simulation. These demonstrations are then adapted to robotic hands and transferred to real robots, including those with different hand designs. Tasks run autonomously, with no teleoperation and no real-robot training data, covering complex scenarios such as object handovers and continuous-contact manipulation.

The DexGEM platform is built on DexRobot's dexterous hardware (DexHand021 series), native perception, and manipulation datasets accumulated over years. These datasets are trained through in-house models and validated in real interaction scenarios, ensuring the reliability and effectiveness of the system. The company's complete dexterous hand portfolio, native multimodal perception, and self-developed data acquisition and models turn human demonstrations into training data and skills, validating every link in the pipeline from hardware to physical tasks.

DexRobot's booth at IROS 2026 highlights the company's commitment to advancing dexterous manipulation technology. The DexHand021 Pro and DexGEM platform demonstrate the company's ability to bridge the gap between AI models and real-world dexterity, setting a new standard in embodied AI.