Humanoid robots are transitioning from research projects to real factory floors, but the industry faces a critical bottleneck: safety standards that can keep pace with AI-driven machines working alongside human workers. At the 2026 Robotics Summit & Expo in Boston, a panel of leading roboticists and engineers revealed that while bipedal robots can now walk, manipulate objects, and learn from real-world data, the lack of standardized safety protocols is slowing mass deployment.

Why Are Humanoid Robots Harder to Deploy Than Robotic Arms?

Robotic arms have been refined for decades in manufacturing, but humanoid robots introduce a fundamentally different challenge. Unlike fixed-position arms, bipedal robots must navigate dynamic environments filled with human workers, forklifts, and unpredictable obstacles. This complexity is why Boston Dynamics, Agility Robotics, and other companies are taking a measured approach to scaling production.

Alberto Rodriguez, director of robot behavior for Atlas at Boston Dynamics, explained that the company's strategy focuses on three interconnected layers: hardware design, behavioral models and architectures, and deployment strategy. "If you fail on finding a general strategy for any one of those three things, it becomes too expensive," he stated. Boston Dynamics has committed to deploying approximately 25,000 humanoids in factories and plans to ramp production capacity to 30,000 Atlas robots per year by 2028.

Agility Robotics has taken a similar path, moving beyond pilot projects with companies including Amazon, GXO, Schaeffler, Toyota, and Mercado Libre. The company is now expanding from container and tote manipulation toward more complex item-level handling tasks.

What's Blocking Humanoid Robots From Scaling Faster?

The biggest obstacle isn't engineering prowess; it's safety validation. Traditional safety methodologies rely on deterministic, predictable behavior. Humanoid robots powered by artificial intelligence introduce learned behaviors that are harder to predict and validate. This creates several engineering challenges that industry standards haven't yet addressed.

• Failure Prediction: Engineers struggle to accurately predict how an AI-driven robot will behave in edge cases or novel situations it hasn't encountered during training.
• Performance Repeatability: Unlike mechanical systems with fixed parameters, AI models can produce slightly different outputs each time, making it difficult to guarantee consistent, safe behavior.
• Human-Robot Risk Assessment: Standards must account for the unique risks of robots working in close proximity to human workers without physical barriers or guards.
• AI Decision Validation: Regulators and manufacturers need methods to verify that an AI system's decisions are safe and explainable, not just accurate.

Aaron Prather, director of the Robotics and Autonomous Systems Program at ASTM International (a standards-setting organization), noted that safety efforts are underway at multiple levels. "Safety efforts are under way at ISO, ASTM, and NIST," he explained. "We are starting to work on the performance repeatability tests, and there are numerous other efforts going on". Prather added that the first safety standards are expected to be drafted within the next two years.

Agility Robotics has taken a proactive step by spinning up an ISO committee and working group with Boston Dynamics and others to study safety issues and develop solutions for the next generation of robots.

How Are Standards Bodies Testing Humanoid Robot Safety?

NIST (National Institute of Standards and Technology) is developing a standardized test bed with approximately 10 tests designed to evaluate humanoid robot capabilities in real-world conditions. These tests will measure locomotion, manipulation, and other critical functions. Some manufacturers will receive test beds for internal use, while others will compete in public robotics events to demonstrate their robots' safety and reliability.

The test bed approach serves a dual purpose: it helps manufacturers validate their designs and provides a common benchmark for developing industry-wide standards. This is particularly important because humanoid robots from different companies will eventually work in the same facilities, requiring interoperable safety protocols.

What Role Does AI Play in Closing the Gap Between Lab and Factory?

Mike Nielsen, chief marketing officer at RealSense (a computer vision company), highlighted that simulation technology is accelerating the development cycle. "The sim2real gap is closing," he noted, referring to the challenge of transferring behaviors learned in simulation to real-world robots. Companies are collaborating with NVIDIA and others to develop universal vision models that can be deployed across different robot platforms in real time, including noise models that account for real-world sensor imperfections.

However, Nielsen emphasized that the biggest remaining gap is applying pure models to real-world conditions. "If you don't have real-world implications built into the model, then you're really just using face and eyeballs," he said. This underscores why companies like Boston Dynamics and Agility are conducting extended factory demonstrations, not just one-off proofs of concept.

Nielsen

Nielsen also noted that product development cycles in China are moving 30 to 40 times faster than in other regions, driven by higher tolerance for risk and rapid iteration. This competitive pressure is pushing companies worldwide to accelerate their own timelines.

Steps to Understanding Humanoid Robot Deployment Timelines

• Pilot Phase: Companies like Boston Dynamics and Agility are currently in extended pilot deployments with major logistics and retail partners, testing real-world performance and gathering safety data.
• Standards Development: Industry bodies are drafting safety standards expected within two years, which will be required before mass deployment can accelerate.
• Production Scaling: Boston Dynamics has committed to 30,000 units per year by 2028, signaling that manufacturers believe safety and technical hurdles are on track to be resolved.
• Regulatory Approval: As standards emerge, manufacturers will need to certify their robots against these benchmarks before deploying them in new facilities or markets.

The humanoid robot industry is at an inflection point. The technology works, companies have secured customer commitments, and production capacity is being built. What's holding back explosive growth is not innovation but validation; the industry must prove to regulators, facility managers, and workers that these machines are safe enough to operate alongside humans at scale. The next 18 to 24 months will be critical in determining whether humanoid robots become a standard part of factory and logistics operations or remain a niche solution for specialized tasks.