NEURA Robotics Launches Gen 4 Humanoid: Each Arm Makes Its Own Millisecond Decisions
NEURA Robotics Launches Gen 4 Humanoid: Each Arm Makes Its Own Millisecond Decisions
NEURA Robotics has launched the Gen 4 humanoid robot, featuring a 'Smart Limbs' architecture where each arm processes sensory data locally for reflexive responses. The robot also incorporates Neuron AI, a Perception-Language-Action model that integrates tactile, force, and audio sensing, addressing limitations in current vision-centric approaches. The TUM RoboGym serves as a data collection infrastructure for training and deploying real-world robotic capabilities.
NEURA Robotics' Gen 4 humanoid introduces a 'Smart Limbs' architecture with distributed edge processors for millisecond-level reflexes, addressing latency issues in real-world factory environments. The Neuron AI model integrates tactile, force, and audio sensing alongside vision and proprioception, marking a shift from vision-centric approaches in industrial robotics.
Source: Robotics — Web news discovery · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
square meters
2,300
View original evidence
at the opening of the TUM RoboGym at Munich Airport — a 2,300-square-meter (24,757 square feet) Physical AI training facility at LabCampus near Munich Airport, jointly developed with the Technical University of Munich (TUM) at a combined investment of approximately €17 million (about $19.1 million USD), with NEURA contributing roughly €11 million (approximately $12.3 million USD). Gen 4 Moves the Robot's Reflexes Int
Open source S3square feet
24,757
View original evidence
at the opening of the TUM RoboGym at Munich Airport — a 2,300-square-meter (24,757 square feet) Physical AI training facility at LabCampus near Munich Airport, jointly developed with the Technical University of Munich (TUM) at a combined investment of approximately €17 million (about $19.1 million USD), with NEURA contributing roughly €11 million (approximately $12.3 million USD). Gen 4 Moves the Robot's Reflexes Int
Open source S3
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
at the opening of the TUM RoboGym at Munich Airport — a 2,300-square-meter (24,757 square feet) Physical AI training facility at LabCampus near Munich Airport, jointly developed with the Technical University of Munich (TUM) at a combined investment of approximately €17 million (about $19.1 million USD), with NEURA contributing roughly €11 million (approximately $12.3 million USD). Gen 4 Moves the Robot's Reflexes Int
Open source S3
, compounded across a full manufacturing shift, produces a system that cannot be relied upon. One benchmark figure illustrates the scale of the problem: one well-regarded humanoid system required half a million real-world hours of training data to reach 99 percent task success rates on a controlled benchmark. Real-world data is not scrappable from the internet. It must be collected physically, by operating real robot
Open source S13
nich become available to a robot handling packaging in Leipzig. Training data flows from real-world operations into Neuron AI model updates, which then flow back to every robot in the network. The TUM RoboGym accelerates that accumulation through three parallel data collection pipelines running simultaneously: robots performing actual industrial tasks while capturing sensor data; human operators wearing full data sui
Open source S18
ts conducting teleoperated demonstrations that generate human-motion training data directly; and a simulation layer that expands the dataset with synthetic variations of documented scenarios. The facility had partner companies from more than ten industries actively working on-site since September 2026 — two months before its official opening — according to NEURA. Sectors include automotive, pharmaceuticals, medical t
Open source S19
echnology, packaging, logistics, agriculture, and care. How Does This Fit in a Race Already Running at Full Speed? By summer 2026, four companies were already deploying humanoids on assembly lines or ramping mass production: Tesla's Optimus (targeting sub-$20,000 per unit), Figure AI's F.03 (operating at BMW facilities, with Figure's factory producing roughly one robot per 90 minutes), Boston Dynamics' Atlas (backed
Open source S20
Source:Robotics — Web news discovery · techtimes.com