Sharpa Robotics

DeepMind’s latest robotics framework debuts three new models, expanding vision-language-action capabilities from isolated tabletop manipulation to full-body humanoid control, high-DoF tactile dexterity, and multi-agent coordination.

Announced at GTC Taipei, the NVIDIA Isaac GR00T Reference Humanoid Robot bundles a Unitree H2 Plus chassis, Jetson Thor compute, and Sharpa Wave tactile hands into a unified open platform.

Sharpa Robotics has unveiled a new hierarchical framework that combines shared-autonomy data collection with a "mixture-of-experts" AI model to achieve complex, contact-rich manipulation tasks.

NVIDIA researchers have revealed EgoScale, a framework that leverages a massive 20,854-hour egocentric human dataset to train robots in complex, fine-grained manipulation with minimal robot-in-the-loop data.

Ahead of CES 2026, the Singapore-based robotics firm details its reliability testing and confirms a modular design for its SharpaWave end-effector.