aiXopt
Back to references

Research cooperation

3 objectives instead of 5 abstract controller gains

How aiXopt makes a path controller tunable through understandable objectives instead of controller parameters only experts understand.

Category
Research cooperation
Transferability
Approach transfers to robots, handling systems and machine axes, including situation-dependent switching between settings.

Challenge

The path controller of an automated system has 5 weighting parameters whose effect on the behavior is hard to predict even for experts. What is actually required are 3 tangible things: path accuracy, speed and smoothness. They conflict, and depending on the situation a different compromise is right. In addition, some settings make the system leave the path, and the trial has to be aborted.

Solution

aiXopt's Small-Data AI does not search for a single setting but for the set of all optimal trade-offs between the 3 objectives, using aborted trials as information along the way. The result is a set of validated settings in which a user moves along the objectives, for example more smoothness instead of speed, without having to understand the controller parameters themselves. With a small trial budget the approach clearly outperformed all compared methods.

More references

Research cooperation

100% robust controller settings in milling

How aiXopt tunes the control loops of an existing milling machine to varying machine behavior and gets the most out of the hardware already in place.

Your machines can do more.
Let's find out how much.

Book a no-obligation call and find out where your biggest potential lies.