Neuroplastic AI for Embodied Control and Automation
We use Neuroplastic AI for real-time adaptive control, delivering the highest intelligence-per-kB AI, and up to 400x greater efficiency than traditional deep learning.
 How can our technology help in your industry Image

Bio-mimetic AI for robotics and industrial process control

We created a revolutionary AI framework which is based on neuroplasticity.

 

We created a revolutionary AI framework which is based on neuroplasticity. We use this technology to make optimal closed loop AI control systems. The neuroplasticity allows the AI to achieve optimal performance by continuously adapting to real-time variations in the equipment, process and environment.

Image

Adaptive Intelligence uses real time feedback to continuously retrain models in the runtime environment based on new data and user adjusted goals. This means it responds quickly to environmental changes, dynamically builds up its own training data, and generates personalised results. Adaptive Intelligence can also respond to challenges that were not foreseen when the model was originally designed or built.

 

This powerful “adaption at the edge” eliminates the need to constantly retrain in the cloud with new data.

Why Adaptive Intelligence

AI offers unprecedented performance in real-time control with a significantly larger operational envelope than traditional control techniques such as PID or MPC, which means much more value for end users.

 

Adaptive Intelligence is different to other AI solutions. The compute efficiency and runtime speed that neuroplasticity delivers allows for very flexible deployment options. It can be installed everywhere, including inside the firmware of old equipment or as a secret algorithm inside a new smart product, or platform.  

Adaptive intelligence runtime offers exceptional robustness, making it perfect for industrial control and optimisation applications. 

 

Our adaptive neural controllers handle variation and continuously achieve optimum performance even when the mechanical equipment wears, or the composition and properties of feedstock changes, or the environmental conditions vary.

Our AI controllers are trained on 1st principles physics models and digital twins of the plant. This means we don’t need the large data sets typically required for AI/ML models.
Instead, we can derive our training model from plant models your engineers already have.

 

Because of both the way the networks are evolved and the live adaption capabilities we can achieve very high computational efficiency, and high data efficiency during training.

These three main key benefits map particularly well to our target customers in industrial applications, because of the limited data, limited compute and continuous variation in the processes.

 

Furthermore, we can package our AI controllers with the physics models to provide advisory systems and co-pilots. This means we offer more than just optimal control to customers. We can offer offline tools to help them understand their plant and equipment better.

Image

Extreme Performance

Image

Powerful
Algorithms

Step-change in performance
Image

Adapts To
Runtime Variation

Robust to system dynamics. Enables autonomous operation

Image

High Clock
Speed

Computes in milliseconds
Image

Modes Of
Operation

Optimise for multiple runtime objectives e.g. performance or energy saving modes

Image
Powerful
Algorithms
Step-change in performance
Image
Adapts To
Runtime Variation

Robust to system dynamics. Enables autonomous operation

Image
High Clock
Speed
Computes in milliseconds
Image
Modes Of
Operation
Optimise for multiple runtime objectives eg. performance or energy saving modes

Flexible Deployment

Image
On-Premise/
On-Board
No cloud or remote connectivity
Image
No Historical
Data Needed

Uses 1 principles model approach and generates own training data

Image
Fits Into
Small Spaces
Very compute efficient
Image
FLEXIBLE
EMBODIMENT

Deploy into embedded controllers, or in PLC/DCS/Motion/SCADA products

Image

On-Premise
/ On-Board

No cloud or remote connectivity
Image

No Historical
Data Needed

Uses 1 principles model approach and generates own training data

Image

Fits Into
Small Spaces

Very compute efficient
Image

Flexible
Embodiment

Deploy into embedded controllers, or in PLC/DCS/Motion/SCADA products

Technology explanation

We would like to help you learn as much as possible about Adaptive Intelligence. To help you understand as quickly as possible, we have written a technical description for control engineers and another one for computer scientists from an artificial intelligence background.

Download the 2 white papers here:

Why AI for control? An engineers perspective
Why AI for control? A computer scientist's perspective

Get in Touch

Schedule a Free Consultation

Industries

Aerospace

Mining & Minerals

Metals

Power Electrification

Marine

Manufacturing

Oil & Gas
Image
Aerospace
Image
Mining & Minerals
Image
Metals
Image
Power Electrification
Image
Marine
Image
Manufacturing
Image
Oil & Gas

Customer example in composite moulding

Surface Generation is an OEM that builds specialised tooling for Injection Moulding and Compression Moulding machines.

 

 

To mould difficult shapes, hi-tech heating elements consisting of smaller heating zones (or “pixels”) are required across the tool face.

The ability to heat and cool each pixel at high speed is a complex non-linear temperature control problem as the temperature of each pixel affects the neighbouring pixels. Achieving strict temperature control across different zones of the same piece of metal requires a control solution that understand the physics of the equipment so that it can adapt in real-time to the machine state and the hundreds of temperature profiles (recipes) that may be presented.

Surface Generation successfully implemented an Adaptive Neural System in both injection moulding and compression moulding machines with significant value improvements in control performance and reduction in energy and compressed air usage.  
  • 10% increase in energy efficiency
  • 33% decrease in compressed air usage
  • 40% improvement in temperature tolerance
Image
Image

Success Stories

Image
Luffy AI is the most credible team on the market.

– Surface Generation CEO

Image
Luffy AI is the most credible team on the market.”

– Surface Generation CEO

Get in Touch

We work with your team to develop system models and a training curriculum that accommodates physical tolerances, safety limits, process scenarios and control uncertainties, while simultaneously and continuously optimising around your value drivers.

Contact Us to Schedule a Consultation

Get in Touch

We work with your team to develop system models and a training curriculum that accommodates physical tolerances, safety limits, process scenarios and control uncertainties, while simultaneously and continuously optimising around your value drivers.

Contact Us to Schedule a Consultation
Scroll to Top