Viral videos of robots kicking children or going out of control in warehouses have become a wake-up call for the entire industry. The main question facing developers today is how to deploy a humanoid machine in a workspace without endangering people. The answer is a multi-layered protection system, ranging from specialized chips to completely abandoning legs in favor of wheels.
The Problem of Probabilistic Systems
Unlike traditional industrial robots (welding machines or palletizers), which are deterministic and follow strict rules, humanoid machines operate based on artificial intelligence. They belong to probabilistic systems, where every action is based on statistical calculations rather than rigid logic. It is this uncertainty that creates the main risk: if a robot loses power, it could simply fall and crush a person, as functional safety experts note.
Hardware and Software Solutions
Nvidia has already proposed a solution based on Blackwell chips — a safety system that interprets sensor data about potential threats and can instantly stop the robot in unsafe conditions. The company's senior director of robotics emphasizes that the functional system and the safety system now work in close conjunction, creating a unified software stack.
Fort Robotics from Philadelphia has gone even further: their controllers collect information not only from the robot's cameras but also from the surrounding infrastructure. This involves complex analysis — where a person is located, in what posture, and whether this data can be trusted for decision-making. This is no longer just visual detection, but full-fledged contextual analysis.
Abandoning Legs and "Collapsing" Falls
Some manufacturers have decided to eliminate the very cause of the risk. Dexmate creates robots on wheeled platforms with long manipulators — the battery and electronics are located in the base, providing a low center of gravity and eliminating falls. Others, like Neura Robotics, are developing bipedal robots that, upon joint failure, attempt to regain balance, and if that is impossible, fall "like a collapsing building," folding downwards to minimize damage.
The International Organization for Standardization (ISO) has already created an expert group to study loss of stability, but unified requirements are not expected until mid-2028. For now, each manufacturer is developing its own safety scenarios.
Expert Perspective
The humanoid robot market is in a phase of "growing pains," and viral incidents are just the tip of the iceberg. The problem is not only technical but also one of trust: without universal safety standards, mass adoption will be hindered. Solutions like wheeled platforms or specialized chips are temporary measures. A real breakthrough will occur when the industry can guarantee that a robot will not pose a threat even in the event of a malfunction, which requires a fundamentally new approach to AI architecture and sensor systems.