Neura Robotics has presented the fourth generation of its 4NE1 humanoid robot in Munich. The 64 kg model is 20% lighter, according to the manufacturer, and has new reactive limbs, improved hands and an architecture designed for series manufacturing.
The German manufacturer introduced the new version at the opening of TUM RoboGym, a robot training centre at the Technical University of Munich (TUM). The new 4NE1 combines a revised mechanical design with Neuron AI, the physical intelligence model developed by Neura Robotics. The company describes this generation as a step towards industrial use and series production of its humanoid.
Neura 4NE1 specs: 1.8 m tall, 64 kg, with a textile covering
Neura Robotics reports a 20% weight reduction compared with the previous generation. The new 4NE1 is 1.8 m tall and weighs around 64 kg, according to the specifications released by the manufacturer.
Its silhouette is slimmer and its appearance has also changed. The robot is now covered in a soft black fabric, unlike the previous version, whose mechanical components had no textile covering.
According to Neura, the covering is meant to make the robot look more approachable and better suited to environments where humans and machines work side by side.
How the Smart Limbs make the 4NE1’s movements more responsive
The fourth-generation 4NE1 features Smart Limbs, limbs that can respond independently to forces from their environment. This architecture is intended to let the robot adjust its movements faster, without always relying on whole-body coordination.
Neura Robotics says this results in faster reactions and more precise, smoother movements. The hands have also been redesigned to improve grip precision compared with the previous model.
These changes are aimed in particular at industrial handling tasks, where the ability to adjust a movement based on contact with an object matters.
Neuron AI: how vision, touch and force data control the robot
The new 4NE1 runs Neuron AI, Neura Robotics’ proprietary physical intelligence model. Its role is to link what the robot perceives to the physical actions it has to carry out.
The system combines several types of data:
- Visual information from the robot’s sensors.
- Tactile data collected during contact.
- Forces applied during handling.
- Sound information from the environment.
- Data describing the robot’s overall state.
This approach is intended, for example, to let the 4NE1 check that assembled components are properly fastened. That kind of check requires taking mechanical forces and contacts into account, not just visual analysis.
According to Neura, Neuron AI combines perception, language understanding and action to turn this information into movement commands.
How Neura Gyms train humanoid robots for industrial tasks
Neura Robotics relies on its training centres, the Neura Gyms, to develop its robots’ practical skills. Training combines real robot data, human demonstrations and simulation.
The skills acquired can then be shared across multiple connected robots. The centres are also used to test these skills in practical situations, with the aim of preparing the humanoids for industrial applications.
The manufacturer says it designed the fourth-generation 4NE1 for series manufacturing. This industrial focus accompanies the presentation of the new model, but it is not an announcement of commercial availability or deliveries.
Three-year, €17 million partnership with the Technical University of Munich
TUM RoboGym is intended to develop and test robot skills that meet companies’ needs. The centre brings researchers and industry players together around a shared research and training infrastructure.
The cooperation between Neura Robotics and the Technical University of Munich is planned to run for three years. The project represents a total investment of €17 million, of which €11 million is funded by Neura Robotics.
Under the partnership, Neura provides robots, technical equipment, computing power and operating staff. TUM provides the premises, its scientific expertise and the resources needed for research activities and for running the centre.
TUM RoboGym is meant to let companies test robotic applications with researchers, then use the results of these trials to guide new scientific work.