Mobile-Menu

Artificial Intelligence Neura Gives Robots a Brain with Neuron AI

From Sariana Kunze | Translated by AI 3 min Reading Time

Related Vendors

FAULHABER Drive Systems
Taiwan Machine Tool & Accessory Builders' Association (TMBA)
SIMTOS (Seoul International Machine Tool Show)

What if robots could perceive more than just what is visible? Neura Robotics introduces Neuron AI: an AI that combines senses, memory, and action—thereby opening up new possibilities for autonomous robots.

Neuron AI combines 3D vision, touch, hearing, proprioception, and language—along with memory—into a Perception-Language-Action (PLA) model. Robots can grasp objects they cannot see.(Image: Neura)
Neuron AI combines 3D vision, touch, hearing, proprioception, and language—along with memory—into a Perception-Language-Action (PLA) model. Robots can grasp objects they cannot see.
(Image: Neura)

Neura Robotics has unveiled Neuron AI, a multimodal foundation model for Physical AI. The AI combines 3D vision, touch, hearing, proprioception, and language with a memory. It enables robots to respond to information that goes beyond what cameras alone can capture—as well as to retain relevant context in memory when objects disappear from view. The Perception-Language-Action (PLA) architecture links what a robot sees, hears, and feels with what it has already done, thereby transforming multisensory information into physical actions.

A Welding Robot Can Tell if the Weld is Successful

Simple cameras capture geometry, color, and position, but do not provide information about touch, force, or sound. People, too, rely on more than just their eyes: They can feel whether a screw is tight and hear when a connector clicks into place. Neuron AI applies this principle to robots. Instead of processing each sensor individually, the model encodes 3D vision, tactile and force data, hearing, the robot’s own state, and speech, and translates them into a continuous sequence of actions.

Multisensory perception is particularly crucial for tasks where cameras alone reach their limits. When a robot unlocks a lock, it is only the clicking sound and the perceptible resistance that indicate it is truly open. With the help of its sense of touch, a robot can grasp objects it cannot see. These additional senses also improve performance on tasks that cameras are already capable of handling: a welding robot, for example, can “hear” whether the weld is successful.

With Neuron AI, we are building the brain of the physical world: Physical AI that transforms human expertise and physical experience into skills that robots can learn and apply. Developed in Germany, built for the world.

David Reger, Founder and CEO of Neura Robotics

Neuron AI Has a Memory

Furthermore, Neuron AI has a memory. To function effectively in everyday life, a robot needs an understanding of cause and effect, as well as the ability to remember objects even when they are out of sight—a skill that humans acquire, for example, during their first two years of life. Unlike purely digital AI, a robot operates in a continuous physical world and must keep track of objects and the consequences of its own actions over time—often without seeing them. Neura has adapted the model’s architecture and technology to incorporate this memory function and is training it using specialized datasets that leverage it.

Neuron AI Is Trained Using Three Data Pillars

Neuron AI is trained using three complementary sources of physical experience:

  • Real-world robot data: Various robots perform tasks in the Neura Gyms—Neura’s training infrastructure for cognitive robots and Physical AI—as well as in real-world environments, generating synchronized data on vision, touch, force, hearing, and motion.
  • Simulation-Enhanced Data: Neuras Data Amplifier and the simulation environment expand the training data by adding simulated variations of tasks and environments.
  • Real human data: People wearing sensor-equipped data suits capture typical human everyday movements, thereby contributing practical expertise to the training pipeline.

Together, these pillars combine real-world physical interactions, simulated variations, and human expertise. This creates a shared data foundation for Neura robots in five form factors, each with different grippers and hands.

Neura Gyms combine data collection, training, and benchmarking prior to launch. Training data and skills developed in the gyms are incorporated into future model releases.

Neuraverse provides the platform to deploy these capabilities across the fleet and update them continuously. It also provides partners with tools to tailor the models to their own tasks, thereby transforming process expertise into robotic capabilities. If desired, partners can also keep their knowledge to themselves. If they decide to share capabilities or operational data, these contributions can support the broader ecosystem and future model development. This makes “time-to-skill” a key metric for progress: how quickly a robot can learn a new task and perform it reliably. For industrial customers, this reduced time is expected to be crucial for integrating Physical AI into more processes and adapting it to changing requirements.

Subscribe to the newsletter now

Don't Miss out on Our Best Content

By clicking on „Subscribe to Newsletter“ I agree to the processing and use of my data according to the consent form (please expand for details) and accept the Terms of Use. For more information, please see our Privacy Policy. The consent declaration relates, among other things, to the sending of editorial newsletters by email and to data matching for marketing purposes with selected advertising partners (e.g., LinkedIn, Google, Meta).

Unfold for details of your consent

Related content