Nvidia Develops Halos Software to Humanize Robot Safety

The company’s new software, Halos, adapts sensor-fusion technology originally designed for self-driving cars to give robots a precise, real-time understanding of their environment. By moving beyond traditional systems that require robots to be caged or stopped, Nvidia aims to enable fluid collaboration. In practical terms, this would allow machines to perform tasks like passing tools or guiding heavy objects alongside human workers without triggering emergency shutdowns at every movement.

Industry leaders argue that this transition from stationary automation to intelligent mobility is essential for the sector’s growth. With Barclays projecting the humanoid robot market to reach $200 billion by 2035, Big Tech firms are racing to provide the full-stack platforms necessary for deployment in factories, hospitals, and homes. As Amit Goel, a senior director at Nvidia, noted, the current standard of simply detecting obstacles and stopping is insufficient for the next generation of physical AI. Instead, the focus is shifting toward systems capable of nuanced, high-stakes decisions, such as a forklift anticipating a human path around a corner rather than reacting only after a collision risk becomes immediate.