Viral videos of humanoid robots kicking children or performing chaotic dances have become a vivid illustration of the industry's main problem: how to deploy a machine in a warehouse or office without putting people at risk. Incidents involving severe injuries and even fatalities have already been recorded, forcing developers to fundamentally rethink safety approaches while simultaneously slowing the pace of mass adoption.
From Determinism to Probabilities: A Fundamental Shift
The key difference between new humanoid robots and traditional industrial machines lies in their architecture. Industrial robots (welders, palletizers) are deterministic systems operating on a rigid set of rules. Humanoids, driven by AI, belong to probabilistic systems: their actions are based on statistical probabilities rather than absolute predictability. This creates the main threat: "If a humanoid robot loses power, it could fall and crush you," describes functional safety expert Michel Silva, outlining the basic risk.
Multi-Layered Protection: From Chips to Abandoning Legs
The industry is responding to the challenge comprehensively. Nvidia has already introduced a safety system for humanoids based on Blackwell chips. The company's Senior Director Amit Goel explained that the model interprets sensor data about potential hazards and can instantly stop the robot in unsafe conditions. "The safety system and the functional system must interact frequently with each other... We created this operating system layer," he noted.
Another level of control involves using external infrastructure. Fort Robotics is developing controllers that gather data from multiple sources: not just detecting a person's presence, but also analyzing their posture, position, and the degree of confidence in this information for decision-making.
The problem of stability loss is so serious that the International Organization for Standardization (ISO) has created a separate expert group. Publication of unified requirements is expected only by mid-2028. Meanwhile, manufacturers are seeking their own solutions. German company Neura Robotics, which produces the 80-kilogram robot 4NE1, states that in the event of a joint failure, the machine will fall "like a collapsing building," folding downward to minimize damage.
The most radical solution was proposed by Dexmate: the company deliberately abandoned legs in favor of wheeled platforms. Co-founder Yuzhe Qin explained that the battery and electronics are placed in the platform, providing a low center of gravity and eliminating the risk of falling. A similar approach is used by Cobot, whose wheeled manipulators move at walking speed and lack super-strong grip. "We don't need to put a lot of energy into necessary actions. We're not trying to crush watermelons," the company's founder Brad Porter quips.
Cryptalist Analysis: The industry finds itself in a paradoxical situation: the smarter and more autonomous robots become, the harder it is to guarantee their safety. The transition to probabilistic AI-based systems is both a breakthrough and a source of new risks that cannot be solved by simply copying industrial standards. Judging by the timeline for publishing ISO standards, the market faces at least 3-4 years of a "Wild West," where only those who can convincingly prove the safety of their machines will survive. Abandoning legs in favor of wheels is not a step backward, but a pragmatic compromise that could become a dominant trend in logistics and warehouses.