The DexGEM platform centers on a technical closed loop for physical AI, detailed in a research paper titled "Dexterous Robot Manipulation from Human Demonstrations via Contact-Anchored Retargeting and Residual Policy Learning." By reconstructing human movements from motion capture and first-person video within a simulation environment, the system transfers complex skills—such as two-hand coordination and continuous-contact manipulation—to physical hardware. This approach bypasses traditional teleoperation, allowing robots to learn nuanced dexterity independently.
DexRobot Debuts DexGEM and DexHand021 Pro at IROS 2026
At the IROS 2026 conference in Pittsburgh, DexRobot is unveiling a new paradigm for embodied AI, headlined by the DexGEM research pipeline and the high-precision DexHand021 Pro. The company is demonstrating how human motion capture translates directly into autonomous, contact-rich robotic manipulation without the need for prior real-world training data.
Supporting this software milestone is the DexHand021 Pro, a 22-degree-of-freedom robotic hand designed for research-grade durability and sensing. The device integrates an actively driven 2-DoF wrist with a dual-tendon bidirectional drive, facilitating precise force control and passive compliance. To manage interaction, the hand features a 3D tactile envelope array capable of 0.01 N normal-force resolution at 200 Hz, alongside a 160-degree wide-angle palm camera. An onboard edge AI chip processes this multimodal data to execute millisecond-level decisions, a capability validated by over 300,000 durability test cycles.
Comments (0)
No comments yet. Be the first!