Nvidia Corp. is developing software aimed at enhancing the safety of humanoid robots operating in close proximity to humans, a move the company says is necessary for robots to be trusted collaborators in workplaces. The technology, announced Monday, builds on Nvidia’s existing systems designed for self-driving vehicles and is intended to improve robots’ real-time awareness and decision-making capabilities.
Humanoid robots—machines designed with human-like features such as heads, torsos, and limbs—are expected to become a significant market within artificial intelligence, with Barclays estimating a $200 billion revenue opportunity by 2035. Nvidia and other Silicon Valley companies are intensifying efforts to capture this growing sector, which envisions robots performing tasks alongside humans rather than being confined to isolated operations.
Currently, safety protocols limit humanoid robots by requiring them to slow down or stop when human contact is possible, which can hinder productivity and complicate cooperative tasks like handing objects or assisting with lifts. According to Amit Goel, senior director of product management at Nvidia, conventional robotics safety measures, such as cages or obstacle detection that triggers a halt, are inadequate for humanoid robots that need to engage in more fluid, interactive behaviors.
Nvidia’s solution centers on Halos, software adapted from its autonomous vehicle technology, paired with the company’s IGX Thor hardware. This platform will enable robots, including products like Agility Robotics’ Digit, to independently analyze their surroundings and make nuanced decisions to safely interact with people. The system also integrates external sensors to provide a comprehensive environmental view—for example, allowing an autonomous forklift to access remote cameras for better situational awareness and adjust its speed accordingly.
The company has established a dedicated lab to assist robot manufacturers in conducting safety tests ahead of regulatory approval, offering engineering support to optimize designs and accelerate certification processes.
Humanoid robots face distinct challenges compared to autonomous vehicles, as they must differentiate between safe and unsafe physical interactions with humans. Pras Velagapudi, chief technology officer of Agility Robotics, emphasized that the robots cannot simply rely on limited strength to avoid harm; they must be capable of exerting sufficient force to perform useful tasks while maintaining advanced safety evaluations of their environment.
Although technical hurdles remain in creating functional and safe humanoid robots, industry experts see a phased approach to deployment. Initial applications are focused on structured sectors like warehouses and logistics, where robots have started to be integrated, including at Toyota Motor Corp.’s manufacturing sites in Canada. From there, the technology is expected to extend to retail, healthcare, and construction, each presenting progressively complex environments but also substantial market potential.
Velagapudi highlighted the sizable opportunity in these areas, noting that while home care robots are a longer-term goal, foundational work in industrial settings will pave the way for broader adoption. The ongoing advancements in robotics safety and autonomy could enable more collaborative and productive human-robot interactions across diverse industries in the years ahead.