HAND DexNex Testbed Improves Dexterous Robots’ Ability to Learn from Users
Outcome/Accomplishment
The NSF-funded Engineering Research Center (ERC) Human AugmentatioN via Dexterity (NSF HAND), headquartered at Northwestern University, has advanced the ability for dexterous robots to learn more quickly and effectively from humans.
Impact/Benefits
HAND is developing versatile, dexterous robotic arms and hands that address a broad range of human, industry, and societal needs. These engineered systems—hands, skills, interfaces, and training materials—will dramatically advance robotic manipulation and its accessibility and usefulness to diverse user populations.
Explanation/Background
Dexterity Nexus (DexNex) is HAND’s flagship testbed for bimanual dexterous robots. The immersive teleoperation interface enables users to see only what the robot sees and feel only what the robot feels. This setup enables the robot to learn autonomous manipulation policies from teleoperation demonstration.
At the 2025 IEEE International Conference on Robotics and Automation, the HAND team successfully demonstrated an initial behavior cloning pipeline for DexNex based on diffusion policy—weighing probabilities of actions instead of requiring explicit instructions or predictions of actions. The team used teleoperation data to train an autonomous “box and blocks” controller to transfer small blocks from one box to another. This success is a first step toward enabling dexterous robots to learn from a wider range of user inputs.
Location
Evanston, Illinoiswebsite
Start Year
Microelectronics and IT
Quantum, Microelectronics, Sensing, and IT
Lead Institution
Core Partners
Fact Sheet
Outcome/Accomplishment
The NSF-funded Engineering Research Center (ERC) Human AugmentatioN via Dexterity (NSF HAND), headquartered at Northwestern University, has advanced the ability for dexterous robots to learn more quickly and effectively from humans.
Location
Evanston, Illinoiswebsite
Start Year
Microelectronics and IT
Quantum, Microelectronics, Sensing, and IT
Lead Institution
Core Partners
Fact Sheet
Impact/benefits
HAND is developing versatile, dexterous robotic arms and hands that address a broad range of human, industry, and societal needs. These engineered systems—hands, skills, interfaces, and training materials—will dramatically advance robotic manipulation and its accessibility and usefulness to diverse user populations.
Explanation/Background
Dexterity Nexus (DexNex) is HAND’s flagship testbed for bimanual dexterous robots. The immersive teleoperation interface enables users to see only what the robot sees and feel only what the robot feels. This setup enables the robot to learn autonomous manipulation policies from teleoperation demonstration.
At the 2025 IEEE International Conference on Robotics and Automation, the HAND team successfully demonstrated an initial behavior cloning pipeline for DexNex based on diffusion policy—weighing probabilities of actions instead of requiring explicit instructions or predictions of actions. The team used teleoperation data to train an autonomous “box and blocks” controller to transfer small blocks from one box to another. This success is a first step toward enabling dexterous robots to learn from a wider range of user inputs.