Project Bonsai enables engineers to add intelligent control to their Simulink models and deploy trained brains (AI agents) on real world physical systems by using machine teaching.
Learn more here: https://blogs.microsoft.com/ai/machine-teaching/
Engineers can connect their existing simulation models with Project Bonsai by using our Simulink Toolbox. We’ve created example models that are showing how to replace a traditional control system with a trained AI controller.
Cartpole
Simulink model of an inverted pendulum and one of the classic examples for reinforcement learning. Teach a brain to balance a pole attached to frictionless cart.
Project Moab
Project Moab is a small balancing robot useful for demonstrating machine teaching on a physical device for Engineers. You can teach a brain how to balance variety of balls by using the Simulink model as training environment and then deploy the trained brain to the physical device. More information on Moab and availability here: https://aka.ms/moab.
Follow these steps to experience the service firsthand.
1) Follow the instructions to sign-up for a Bonsai account: https://docs.microsoft.com/en-us/bonsai/guides/account-setup
2) Download and install the toolbox https://aka.ms/as/bonsai-toolbox
3) Decide which example you would like to use
4) Follow the instructions outlined in the readme.md that is part of example
5) Start training a Bonsai brain
Project Bonsai documentation: https://aka.ms/bonsai-docs
Support and feedback: http://aka.ms/as/forums