TARS Built a Robot Hand That Can Feel! 🤯 | ICRA 2026
KhanFlicks host reports from the TARS booth at ICRA 2026 in Vienna, where TARS engineers demonstrate tactile robotics, embodied AI, and autonomous task performance.
In this talk
- 1:08
A hand designed around human motion
TARS presents its Dex Hand as an integrated system of hand, sensors, AI, and training data. Its human-inspired design has 21 degrees of freedom, including complex thumb movement, and can reproduce 26 sign-language gestures. The host demonstrates gesture imitation and notes that the movements respond quickly.
- 3:36
Sensing touch and anticipating change
TARS says fingertip sensors register surface detail as small as 0.05 millimeters, along with pressure and force direction. An engineer shows live sensor readings, including force vectors and signals from multiple fingers. The company combines wrist-force sensing, Tact Foresight, and its AWE 3.0 model to anticipate changes during manipulation.
- 5:42
Completing tasks with two hands
A robot autonomously packs objects into a pencil case, closes it, places it in a backpack, and zips the bag. It uses both arms, adjusts its actions, and retries when needed. TARS says it trains robots to carry out tasks from human demonstrations instead of specifying every coordinate by hand.
- 8:30
Training for precise assembly
TARS says its A1 robot achieved submillimeter wire-harness assembly using AWE 3.0. The company also describes collecting human movement data with a Sense Hub suit, so robots can learn from demonstrations rather than relying only on manually programmed actions. The goal is to help robots adapt beyond controlled work settings.
Briefs summarise talks and interviews; timestamps open the video at that moment. Claims in a talk are the speaker's own.