Sharpa reported new research developed in collaboration with NVIDIA aimed at improving simulation methods used to train robots performing complex manipulation tasks. The work focuses on addressing limitations in existing tactile simulation systems used in robotics training. Sharpa trains robots either through reinforcement learning in simulated environments or by generating synthetic data to pre-train its Vision Tactile Language Model (VTLA). Traditional tactile simulations often require a trade-off between physical accuracy and computational speed.
The companies developed a simulation framework called Tacmap designed to address this constraint through a shared geometric representation that maintains fidelity while allowing faster computation. Sharpa indicated that the simulation framework and associated code assets will be released as open source.
“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 vice president of go-to-market and president of Europe at Sharpa.
Sharpa also stated that its dexterous robotic hand, Wave, has been used by researchers at NVIDIA’s GEAR Lab to explore data-efficient robot learning. The research involved transferring policies derived from pre-training the GR00T model on more than 20,000 hours of human video data to robots equipped with the Wave hands.
According to the companies, robots trained with the model were able to complete tasks including assembling model cars, operating syringes and sorting cards with a reported 54% higher success rate compared with baseline approaches.
Sharpa develops robotic systems designed for manipulation tasks and general-purpose applications. Its portfolio includes North, a humanoid robot platform combining whole-body control and fine manipulation; Wave, a robotic hand with 22 active degrees of freedom and integrated tactile sensors; and CraftNet, a VTLA model designed to coordinate motion and physical interaction with objects.
Sharpa plans to demonstrate its robotic systems and related software at NVIDIA GTC 2026. The company is also participating in the NVIDIA Inception program, which supports technology startups working with NVIDIA platforms.
