Automated PID tuning using Reinforcement Learning

I had an idea about automated PID tuning process which involves neural networks and reinforcement learning.

I want to develop new system which takes into account DIY drone makers' problem with balancing. PID algorithm really makes that but there is big problem with general rules of tuning and balancing.You have to try different variant and maybe you can't get to optimal solution.

I think that unsupervised learning techniques can change the way we are trying to fine tune quads. Suppose we have a training environment and new processor which helps us to fly our DIY drones some time and it balances itself.

Personally I have made quad frame by hand and have problems with balancing my drone, it doesn't have good balance (small differences in motor to motor distances, and controller isn't located directly in the center of mass). I think that would be good idea to make a mechanism which will have ability to solve any balance problems.

I need a help from professionals and please share me your opinions about that.

Views: 370


Reply to This

© 2020   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service