Adaptive control

Better industrial and robotic control

Our novel controller uses machine learning and artificial intelligence methods to improve standard industrial control, including robotics.

These patented techniques build on decades of research from MIT into efficient methods for building robust adaptive controllers for nonlinear systems.

Regardless of whether you’re using position, force, or any other control target, our addition of a neural network into the controller is mathematically guaranteed to improve the control outcome compared to even a well-tuned PID controller.

If desired, the controller can continue to adapt while in use to account for unknown or expected deviations from original tuning conditions, whether that’s because of wear and tear on the plant, or a change in environmental dynamics.

Best of all, we can implement the entire controller on state-of-the-art, efficient neuromorphic hardware (or standard hardware, if you already have it).

Adaptive control Features

Online learning
Online learning
Self-tuning controller
Self-tuning controller
Neuromorphic or conventional
Neuromorphic or conventional
Guaranteed convergence
Guaranteed convergence
Contact us for more information

Demonstration

When applied to a Jaco2 robotic arm, the adaptive controller can compensate for unknown forces and use novel tools.

Follow us

The core algorithms are publicly available in the ABR Control Github repository. Explore the code and follow the project!

Partnerships

If you're not sure this will work for your application, contact us and we'll help you decide. We can work closely with you to apply the controller to your application.

Services

An adaptive controller for your problem

Neuromorphic applications

Let us help with your control problems by developing neuromorphic applications for you. We've built our reputation on developing proof-of-concept applications of neuromorphic technology.

More about Neuromorphic applications >

Research

Robotic Arm

To effectively implement force controllers, accurate models of the system dynamics are required. Nonlinear adaptive control methods provide means of compensating for inaccurate models online, and are provably stable and at least as effective as standard PID control. Our research has shown how nonlinear adaptive control can be implemented in neural networks to control robotic simulations. We applied this control on the Kinova Jaco2 robotic arm by first developing a force-control interface to the arm (available publicly on Github), and then building a hybrid implementation of adaptive neural control using Nengo (which we have also made available publicly) and neuromorphic hardware. Our demonstration shows the system adapting to unexpected forces acting upon the arm in real-time, showcasing the utility of low-power, adaptive, embedded robotic systems.