Home
News
Tech Grid
Interviews
Anecdotes
Think Stack
Press Releases
Articles
  • Robotics & Automation

Sharpa NVIDIA Advance Dexterous Robot Simulation Training


Sharpa NVIDIA Advance Dexterous Robot Simulation Training
  • by: Source Logo
  • |
  • March 16, 2026

Sharpa, an AI robotics unicorn, has unveiled new research in partnership with NVIDIA that significantly improves simulation techniques for training dexterous robots. The collaboration introduces Tacmap, a high-fidelity simulation framework that balances physical accuracy and computational efficiency, enabling better training for complex manipulation tasks in consumer and enterprise applications.

Quick Intel

  • Sharpa and NVIDIA developed Tacmap, a simulation framework using shared geometric representation to overcome trade-offs between realism and speed in tactile simulation.
  • Tacmap and related code will be open sourced to benefit the broader robotics community.
  • NVIDIA’s GEAR Lab transferred policies from its GR00T model (pre-trained on 20,000+ hours of human videos) to robots with Sharpa’s Wave hands, achieving 54% higher task success rates.
  • Demonstrated tasks include assembling model cars, operating syringes, and sorting cards using highly anthropomorphic hands.
  • Sharpa’s Wave hand features 22 active degrees of freedom and integrated tactile sensors; it was selected by NVIDIA for data-efficient learning research.
  • Sharpa joins the NVIDIA Inception program and will showcase its full-stack dexterity solutions, including North humanoid robot and CraftNet VTLA model, at GTC 2026.

“This collaboration strengthens the foundation for training in simulation, advancing the robotics field towards more dexterity and autonomy and accelerating large-scale deployment,” said Alicia Veneziani, Global VP of Go-To-Market and President of Europe at Sharpa.

Traditional simulation methods for robot training often compromise between physical authenticity and speed, limiting progress in achieving human-like dexterity. Sharpa addresses this through Tacmap, developed jointly with NVIDIA, which provides a unified, high-fidelity geometric representation for both simulation and real-world transfer.

Sharpa also highlights successful sim-to-real transfer using its Wave robotic hand. NVIDIA researchers pre-trained the GR00T foundation model on extensive human video data and applied the resulting policies to robots equipped with Wave hands, resulting in substantial performance gains on precise manipulation tasks.

Showcasing Dexterity-First Robotics at GTC 2026

Sharpa’s portfolio includes:

  • North: a general-purpose humanoid robot with whole-body control and fine loco-manipulation capabilities.
  • Wave: a human-scale robotic hand with 22 active degrees of freedom and advanced tactile sensors.
  • CraftNet: a Vision Tactile Language Model featuring a Motion Brain (System 1) for movement direction and an Interaction Brain (System 0) for precise object contact, optimized by mapping tactile data to human video and glove inputs.

Sharpa, recipient of the iF Product Design Award and CES Innovation 2026 Award, will demonstrate these technologies at GTC 2026 (Booth #1838, Hall 3, San Jose McEnery Convention Center). The company is also offering media interviews, product demos, and office tours at its Mountain View location upon request.

Sharpa’s membership in the NVIDIA Inception program further supports its mission to accelerate productive, dexterous robot deployment.

About Sharpa

Founded in 2024, Sharpa is a unicorn AI robotics company dedicated to developing high-performance dexterous robots, systems and components for general-purpose applications. Sharpa's mission is to build robots that free people from repetitive or strenuous work to focus on more meaningful pursuits. Sharpa's global headquarters is in Singapore, with manufacturing R&D in Shanghai and business operations in Mountain View, USA.

  • RoboticsAI DexterityHumanoid Robots
News Disclaimer
  • Share