Exploring Microsoft Machine Teaching Online Service for Building Autonomous Systems Using Simulink Models Highlights
Cyrill Glockner, Microsoft
Microsoft® Machine Teaching Online Service uses techniques such as curriculum learning, deep reinforcement learning, scalable data generation, and others to support subject matter experts who are building intelligent control systems for a wide range of applications.
Published: 14 Dec 2021
[MUSIC PLAYING]
I will talk about autonomous systems and how we as Microsoft have taken I would call it a new approach on how to build them. These initial industrial systems were all manually controlled. When we are now thinking about autonomy in the future, we want to solve previously unsolvable problems. And we want to introduce more robust and flexible decision making. We want to be able to bring in the human in the loop and scale human expertise.
The projects that we are driving are always having some kind of a tie-in into the real world. What we have done here is we took a Simulink based model of the CNC machine. And we taught an AI by doing an interaction with that simulation model over and over again until the AI agent had learned to do an optimal calibration task. I'm using Simulink as the simulation environment for training an AI agent. The end results were that we can do it in 13 seconds instead of two hours with the higher precision than the human would be doing.
We want to make sure that we can take the human expertise that exist. And enable the human to describe the problem to the machine. So you don't need to be a data scientist to be able to solve this class of problems. So the end result will always be a trained neural net that can be connected to the real machine.
[MUSIC PLAYING]