Extra robotic thumb lets wearers complete two-handed tasks with one hand
Overview
Researchers at the University of California San Diego developed EgoThumb, a wearable robotic thumb system that combines an additional robotic thumb, a palm-facing camera, and forearm-mounted motors. The system uses an AI model trained via imitation learning to autonomously plan thumb movements based on visual input, allowing users to perform tasks like unscrewing bottle caps or stirring tea with one hand. The hardware designs and software are open source.
Generated from attributed reports · Updated 2 days ago
Event evidence and corrections
1 attributed source owners. Ownership does not establish independent confirmation. Quantities are reported separately and are never added together.
Environment: controlled real
“ns, without requiring the user to issue explicit commands." Fu and his colleagues developed an initial prototype of their robotic thumb. In this prototype, the AI model analyzing camera images and planning future movements runs on an external computer that is connected to the robotic thumb. The researchers evaluated EgoThumb in laboratory experiments in which three wearers attempted four manipulation tasks. These inc”
Exact source · revision 2science-x
Control: reported autonomous
“ble robotic finger could help hold objects while leaving my own fingers free to work." Credit: Jianyu Hou et al Enhancing one-hand manipulation with an extra thumb Fu and his colleagues set out to develop a compact, wearable robotic system that could serve as an extra thumb to hold objects in place while a wearer's fingers manipulate them. In addition, they wanted the thumb to autonomously determine when to grasp and”
Exact source · revision 2science-x
Reported quantity · Reported success rate: 67 percent · unique elapsed hours · Differing source assertions
“same areas used during training, average success across four tasks was 90% for the demonstrator and 92.5% and 87.5% for two new users." In a separate tea-stirring test, success rates averaged approximately 67% and 73% across three users in two previously unseen settings, respectively. The researchers found that performance declined on some tasks when they expanded the areas where objects were initially found or had”
Exact source · revision 2science-x
Reported quantity · Reported success rate: 73 percent · unique elapsed hours · Differing source assertions
“same areas used during training, average success across four tasks was 90% for the demonstrator and 92.5% and 87.5% for two new users." In a separate tea-stirring test, success rates averaged approximately 67% and 73% across three users in two previously unseen settings, respectively. The researchers found that performance declined on some tasks when they expanded the areas where objects were initially found or had”
Exact source · revision 2science-x
Reported quantity · trials per task per user: 10 other · reported trials
“o new wearers and the initial demonstrator achieved comparable success rates when objects were placed within the same areas the AI algorithm encountered during training. "The most encouraging finding is that EgoThumb can learn to coordinate with a human hand without requiring the user to explicitly command each grasp," said Fu. "In our initial evaluation, with 10 trials per task per user and objects placed within the”
Exact source · revision 2science-x
Reported quantity · tasks evaluated: 4 other · Basis not reported
“ns, without requiring the user to issue explicit commands." Fu and his colleagues developed an initial prototype of their robotic thumb. In this prototype, the AI model analyzing camera images and planning future movements runs on an external computer that is connected to the robotic thumb. The researchers evaluated EgoThumb in laboratory experiments in which three wearers attempted four manipulation tasks. These inc”
Exact source · revision 2science-x
Dataset · modalities: camera images
“mages of the user's hand, the robotic thumb and nearby objects. "We trained an AI model to learn from human demonstrations, an approach called imitation learning," explained Fu. "To collect the training demonstrations, a person controlled the robotic thumb through a foot-operated interface. Once trained, the model predicts the robotic thumb's movements directly from camera images and the thumb's current joint positio”
Exact source · revision 2science-x
Dataset · tasks: picking up a cube · Differing source assertions
“luded picking up a cube, unscrewing a bottle cap, stirring tea and transferring a ball from one location to another. The extra thumb helped wearers complete each type of task using only one hand. EgoThumb demonstrates grasping and holding assistance across four tasks, shown from left to right: picking up a cube, unscrewing a bottle cap, stirring tea and transferring a ball. Credit: EgoThumb research team (Jianyu Hou,”
Exact source · revision 2science-x
Dataset · tasks: unscrewing a bottle cap · Differing source assertions
“luded picking up a cube, unscrewing a bottle cap, stirring tea and transferring a ball from one location to another. The extra thumb helped wearers complete each type of task using only one hand. EgoThumb demonstrates grasping and holding assistance across four tasks, shown from left to right: picking up a cube, unscrewing a bottle cap, stirring tea and transferring a ball. Credit: EgoThumb research team (Jianyu Hou,”
Exact source · revision 2science-x
Dataset · tasks: stirring tea · Differing source assertions
“luded picking up a cube, unscrewing a bottle cap, stirring tea and transferring a ball from one location to another. The extra thumb helped wearers complete each type of task using only one hand. EgoThumb demonstrates grasping and holding assistance across four tasks, shown from left to right: picking up a cube, unscrewing a bottle cap, stirring tea and transferring a ball. Credit: EgoThumb research team (Jianyu Hou,”
Exact source · revision 2science-x
Dataset · tasks: transferring a ball · Differing source assertions
“luded picking up a cube, unscrewing a bottle cap, stirring tea and transferring a ball from one location to another. The extra thumb helped wearers complete each type of task using only one hand. EgoThumb demonstrates grasping and holding assistance across four tasks, shown from left to right: picking up a cube, unscrewing a bottle cap, stirring tea and transferring a ball. Credit: EgoThumb research team (Jianyu Hou,”
Exact source · revision 2science-x
Report timeline
Follow attributed reports and material updates.
- Tech Xplore — RoboticsSignalExtra robotic thumb lets wearers complete two-handed tasks with one hand
Researchers at the University of California San Diego developed EgoThumb, a wearable robotic thumb system that combines an additional robotic thumb, a palm-facing camera, and forearm-mounted motors. The system uses an AI model trained via imitation learning to autonomously plan thumb movements based on visual input, allowing users to perform tasks like unscrewing bottle caps or stirring tea with one hand. The hardware designs and software are open source.
Event coverage history
Current attention 6·Peak within the comparable range 10(2026-11-10 15:00 UTC)·Change within the comparable range over 24 hours –
The trend compares participants observed continuously; its scope may be narrower than current attention totals. Hover or click to explore hourly observations, or use the left and right arrow keys.
Timezone · UTC
Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).